Proc Capability - Sas User Guide

®
SAS/QC 15.3
User’s Guide
The CAPABILITY
Procedure
®
SAS Documentation
January 31, 2023
This document is an individual chapter from SAS/QC® 15.3 User’s Guide.
The correct bibliographic citation for this manual is as follows: SAS Institute Inc. 2023. SAS/QC® 15.3 User’s Guide. Cary, NC:
SAS Institute Inc.
SAS/QC® 15.3 User’s Guide
Copyright © 2023, SAS Institute Inc., Cary, NC, USA
All Rights Reserved. Produced in the United States of America.
For a hard-copy book: No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by
any means, electronic, mechanical, photocopying, or otherwise, without the prior written permission of the publisher, SAS Institute
Inc.
For a web download or e-book: Your use of this publication shall be governed by the terms established by the vendor at the time
you acquire this publication.
The scanning, uploading, and distribution of this book via the internet or any other means without the permission of the publisher is
illegal and punishable by law. Please purchase only authorized electronic editions and do not participate in or encourage electronic
piracy of copyrighted materials. Your support of others’ rights is appreciated.
January 2023
SAS® and all other SAS Institute Inc. product or service names are registered trademarks or trademarks of SAS Institute Inc. in the
USA and other countries. ® indicates USA registration.
Other brand and product names are trademarks of their respective companies.
SAS software may be provided with certain third-party software, including but not limited to open source software, which is licensed
under its applicable third-party software license agreement. For license information about third-party software distributed with SAS
software, refer to Third-Party Software Reference | SAS Support.
Chapter 6
The CAPABILITY Procedure
Contents
Introduction: CAPABILITY Procedure . . . . . . . . . . . . . . . . . . . . . . . . . . . . 203
Learning about the CAPABILITY Procedure . . . . . . . . . . . . . . . . . . . . . . 205
PROC CAPABILITY and General Statements . . . . . . . . . . . . . . . . . . . . . . . . . 205
Overview: CAPABILITY Procedure . . . . . . . . . . . . . . . . . . . . . . . . . . 205
Getting Started: CAPABILITY Procedure . . . . . . . . . . . . . . . . . . . . . . . . 207
Computing Descriptive Statistics . . . . . . . . . . . . . . . . . . . . . . . . 207
Computing Capability Indices . . . . . . . . . . . . . . . . . . . . . . . . . 209
Syntax: CAPABILITY Procedure . . . . . . . . . . . . . . . . . . . . . . . . . . . . 211
PROC CAPABILITY Statement . . . . . . . . . . . . . . . . . . . . . . . . 212
BY Statement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 222
CLASS Statement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 223
FREQ Statement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 225
ID Statement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 225
SPEC Statement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 225
VAR Statement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 229
WEIGHT Statement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 229
Graphical Enhancement Statements . . . . . . . . . . . . . . . . . . . . . . 230
Details: CAPABILITY Procedure . . . . . . . . . . . . . . . . . . . . . . . . . . . . 230
Input Data Sets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 230
Output Data Set . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 232
Descriptive Statistics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 235
Signed Rank Statistic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 237
Tests for Normality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 238
Percentile Computations . . . . . . . . . . . . . . . . . . . . . . . . . . . . 240
Robust Estimators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 243
Computing the Mode . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 245
Assumptions and Terminology for Capability Indices . . . . . . . . . . . . . 246
Standard Capability Indices . . . . . . . . . . . . . . . . . . . . . . . . . . 246
Specialized Capability Indices . . . . . . . . . . . . . . . . . . . . . . . . . 250
Missing Values . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 257
ODS Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 257
Examples: CAPABILITY Procedure . . . . . . . . . . . . . . . . . . . . . . . . . . 260
Example 6.1: Reading Specification Limits . . . . . . . . . . . . . . . . . . 260
Example 6.2: Enhancing Reference Lines . . . . . . . . . . . . . . . . . . . 262
Example 6.3: Displaying a Confidence Interval for Cpk . . . . . . . . . . . . 264
CDFPLOT Statement: CAPABILITY Procedure . . . . . . . . . . . . . . . . . . . . . . . 266
200 F Chapter 6: The CAPABILITY Procedure
Overview: CDFPLOT Statement . . . . . . . . . . . . . . . . . . . . . . . . . . . . 266

Getting Started: CDFPLOT Statement . . . . . . . . . . . . . . . . . . . . . . . . . . 267
Creating a Cumulative Distribution Plot . . . . . . . . . . . . . . . . . . . . 267
Syntax: CDFPLOT Statement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 268
Summary of Options . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 269
Dictionary of Options . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 274
Details: CDFPLOT Statement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 282
ODS Graphics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 282
Examples: CDFPLOT Statement . . . . . . . . . . . . . . . . . . . . . . . . . . . . 282
Example 6.4: Fitting a Normal Distribution . . . . . . . . . . . . . . . . . . 282
Example 6.5: Using Reference Lines with CDF Plots . . . . . . . . . . . . . 284
COMPHISTOGRAM Statement: CAPABILITY Procedure . . . . . . . . . . . . . . . . . . 285
Overview: COMPHISTOGRAM Statement . . . . . . . . . . . . . . . . . . . . . . . 285
Getting Started: COMPHISTOGRAM Statement . . . . . . . . . . . . . . . . . . . . 286
Creating a One-Way Comparative Histogram . . . . . . . . . . . . . . . . . 286
Adding Fitted Normal Curves to a Comparative Histogram . . . . . . . . . . 288
Syntax: COMPHISTOGRAM Statement . . . . . . . . . . . . . . . . . . . . . . . . 290
Details: COMPHISTOGRAM Statement . . . . . . . . . . . . . . . . . . . . . . . . 305
ODS Graphics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 305
Examples: COMPHISTOGRAM Statement . . . . . . . . . . . . . . . . . . . . . . . 306
Example 6.6: Adding Insets with Descriptive Statistics . . . . . . . . . . . . 306
Example 6.7: Creating a Two-Way Comparative Histogram . . . . . . . . . . 308
HISTOGRAM Statement: CAPABILITY Procedure . . . . . . . . . . . . . . . . . . . . . 311
Overview: HISTOGRAM Statement . . . . . . . . . . . . . . . . . . . . . . . . . . 311
Getting Started: HISTOGRAM Statement . . . . . . . . . . . . . . . . . . . . . . . 312
Creating a Histogram with Specification Limits . . . . . . . . . . . . . . . . 312
Adding a Normal Curve to the Histogram . . . . . . . . . . . . . . . . . . . 313
Customizing a Histogram . . . . . . . . . . . . . . . . . . . . . . . . . . . . 316
Syntax: HISTOGRAM Statement . . . . . . . . . . . . . . . . . . . . . . . . . . . . 318
Details: HISTOGRAM Statement . . . . . . . . . . . . . . . . . . . . . . . . . . . . 349
Formulas for Fitted Curves . . . . . . . . . . . . . . . . . . . . . . . . . . . 349
Kernel Density Estimates . . . . . . . . . . . . . . . . . . . . . . . . . . . . 360
Printed Output . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 362
Output Data Sets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 369
ODS Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 372
ODS Graphics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 373
SYMBOL and PATTERN Statement Options . . . . . . . . . . . . . . . . . 373
Examples: HISTOGRAM Statement . . . . . . . . . . . . . . . . . . . . . . . . . . 376
Example 6.8: Fitting a Beta Curve . . . . . . . . . . . . . . . . . . . . . . . 376
Example 6.9: Fitting Lognormal, Weibull, and Gamma Curves . . . . . . . . 379
The CAPABILITY Procedure F 201
Example 6.10: Comparing Goodness-of-Fit Tests . . . . . . . . . . . . . . . 385

Example 6.11: Computing Capability Indices for Nonnormal Distributions . . 387
Example 6.12: Computing Kernel Density Estimates . . . . . . . . . . . . . 388
Example 6.13: Fitting a Three-Parameter Lognormal Curve . . . . . . . . . . 390
Example 6.14: Annotating a Folded Normal Curve . . . . . . . . . . . . . . 392
INSET Statement: CAPABILITY Procedure . . . . . . . . . . . . . . . . . . . . . . . . . . 398
Overview: INSET Statement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 398
Getting Started: INSET Statement . . . . . . . . . . . . . . . . . . . . . . . . . . . . 399
Displaying Summary Statistics on a Histogram . . . . . . . . . . . . . . . . 399
Formatting Values and Customizing Labels . . . . . . . . . . . . . . . . . . 400
Adding a Header and Positioning the Inset . . . . . . . . . . . . . . . . . . . 402
Syntax: INSET Statement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 403
Summary of INSET Keywords . . . . . . . . . . . . . . . . . . . . . . . . . 405
Details: INSET Statement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 418
Positioning the Inset Using Compass Points . . . . . . . . . . . . . . . . . . 418
Positioning the Inset in the Margins . . . . . . . . . . . . . . . . . . . . . . 419
Positioning the Inset Using Coordinates . . . . . . . . . . . . . . . . . . . . 420
Examples: INSET Statement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 423
Example 6.15: Inset for Goodness-of-Fit Statistics . . . . . . . . . . . . . . 423
Example 6.16: Inset for Areas Under a Fitted Curve . . . . . . . . . . . . . . 424
INTERVALS Statement: CAPABILITY Procedure . . . . . . . . . . . . . . . . . . . . . . 426
Overview: INTERVALS Statement . . . . . . . . . . . . . . . . . . . . . . . . . . . 426
Getting Started: INTERVALS Statement . . . . . . . . . . . . . . . . . . . . . . . . 427
Computing Statistical Intervals . . . . . . . . . . . . . . . . . . . . . . . . . 427
Syntax: INTERVALS Statement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 429
Details: INTERVALS Statement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 432
Methods for Computing Statistical Intervals . . . . . . . . . . . . . . . . . . 432
OUTINTERVALS= Data Set . . . . . . . . . . . . . . . . . . . . . . . . . . 435
ODS Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 436
Examples: INTERVALS Statement . . . . . . . . . . . . . . . . . . . . . . . . . . . 436
Example 6.17: Computing One-Sided Lower Prediction Limits . . . . . . . . 436
Example 6.18: Computing Nonparametric Tolerance Limits . . . . . . . . . 437
OUTPUT Statement: CAPABILITY Procedure . . . . . . . . . . . . . . . . . . . . . . . . 439
Overview: OUTPUT Statement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 439
Getting Started: OUTPUT Statement . . . . . . . . . . . . . . . . . . . . . . . . . . 439
Saving Summary Statistics in an Output Data Set . . . . . . . . . . . . . . . 439
Saving Percentiles in an Output Data Set . . . . . . . . . . . . . . . . . . . . 441
Syntax: OUTPUT Statement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 442
Details: OUTPUT Statement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 449
OUT= Data Set . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 449
Examples: OUTPUT Statement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 449

Example 6.19: Computing Nonstandard Capability Indices . . . . . . . . . . 449
Example 6.20: Approximate Confidence Limits for Cpk . . . . . . . . . . . 452
PPPLOT Statement: CAPABILITY Procedure . . . . . . . . . . . . . . . . . . . . . . . . . 454
Overview: PPPLOT Statement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 454
Getting Started: PPPLOT Statement . . . . . . . . . . . . . . . . . . . . . . . . . . . 455
Creating a Normal Probability-Probability Plot . . . . . . . . . . . . . . . . 455
Syntax: PPPLOT Statement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 457
Details: PPPLOT Statement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 470
Construction and Interpretation of P-P Plots . . . . . . . . . . . . . . . . . . 470
Comparison of P-P Plots and Q-Q Plots . . . . . . . . . . . . . . . . . . . . 473
Summary of Theoretical Distributions . . . . . . . . . . . . . . . . . . . . . 474
Specification of Symbol Markers . . . . . . . . . . . . . . . . . . . . . . . . 476
Specification of the Distribution Reference Line . . . . . . . . . . . . . . . . 476
ODS Graphics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 476
PROBPLOT Statement: CAPABILITY Procedure . . . . . . . . . . . . . . . . . . . . . . . 477
Overview: PROBPLOT Statement . . . . . . . . . . . . . . . . . . . . . . . . . . . . 477
Getting Started: PROBPLOT Statement . . . . . . . . . . . . . . . . . . . . . . . . . 478
Creating a Normal Probability Plot . . . . . . . . . . . . . . . . . . . . . . . 478
Creating Lognormal Probability Plots . . . . . . . . . . . . . . . . . . . . . 479
Syntax: PROBPLOT Statement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 483
Details: PROBPLOT Statement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 501
SYMBOL Statement Options . . . . . . . . . . . . . . . . . . . . . . . . . . 503
ODS Graphics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 504
Examples: PROBPLOT Statement . . . . . . . . . . . . . . . . . . . . . . . . . . . . 505
Example 6.21: Displaying a Normal Reference Line . . . . . . . . . . . . . 505
Example 6.22: Displaying a Lognormal Reference Line . . . . . . . . . . . . 506
QQPLOT Statement: CAPABILITY Procedure . . . . . . . . . . . . . . . . . . . . . . . . 508
Overview: QQPLOT Statement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 508
Getting Started: QQPLOT Statement . . . . . . . . . . . . . . . . . . . . . . . . . . 509
Creating a Normal Quantile-Quantile Plot . . . . . . . . . . . . . . . . . . . 509
Adding a Distribution Reference Line . . . . . . . . . . . . . . . . . . . . . 511
Syntax: QQPLOT Statement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 512
Details: QQPLOT Statement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 532
Construction of Quantile-Quantile and Probability Plots . . . . . . . . . . . 532
Interpretation of Quantile-Quantile and Probability Plots . . . . . . . . . . . 533
Introduction: CAPABILITY Procedure F 203
Graphical Estimation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 535

SYMBOL Statement Options . . . . . . . . . . . . . . . . . . . . . . . . . . 537
ODS Graphics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 538
Examples: QQPLOT Statement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 539
Example 6.23: Interpreting a Normal Q-Q Plot of Nonnormal Data . . . . . . 539
Example 6.24: Estimating Parameters from Lognormal Plots . . . . . . . . . 540
Example 6.25: Comparing Weibull Q-Q Plots . . . . . . . . . . . . . . . . . 546
Example 6.26: Estimating Cpk from a Normal Q-Q Plot . . . . . . . . . . . 548
Dictionary of Common Options: CAPABILITY Procedure . . . . . . . . . . . . . . . . . . 550
General Options . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 550
Options for Traditional Graphics . . . . . . . . . . . . . . . . . . . . . . . . . . . . 555
Options for Legacy Line Printer Charts . . . . . . . . . . . . . . . . . . . . . . . . . 558
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 558
Introduction: CAPABILITY Procedure

A process capability analysis compares the distribution of output from an in-control process to its specification
limits to determine the consistency with which the specifications can be met. The CAPABILITY procedure
provides the following:
process capability indices, such as Cp and Cpk
descriptive statistics based on moments, including skewness and kurtosis. Other descriptive information
provided includes quantiles or percentiles (such as the median), frequency tables, and details on extreme
values.
histograms. Optionally, these can be superimposed with specification limits, fitted probability density
curves for various distributions, and kernel density estimates.
cumulative distribution function plots (cdf plots). Optionally, these can be superimposed with specifi-
cation limits and probability distribution curves for various distributions.
quantile-quantile plots (Q-Q plots), probability plots, and probability-probability plots (P-P plots).
These plots facilitate the comparison of a data distribution with various theoretical distributions.
Optionally, Q-Q plots and probability plots can be superimposed with specification limits.
comparative histograms, cdf plots, Q=Q plots, probability plots, and P-P plots. These are composite
graphs that are composed of plots that correspond to the different levels of specified CLASS variables.
goodness-of-fit tests for a variety of distributions including the normal. The assumption of normality is
critical to the interpretation of capability indices.
statistical intervals (prediction, tolerance, and confidence intervals) for a normal population
the ability to produce plots either as traditional graphics, ODS Graphics output, or legacy line printer
plots. Traditional graphics can be saved, replayed, and annotated.
the ability to inset summary statistics and capability indices in graphical output
the ability to analyze data sets with a frequency variable
the ability to read specification limits from a data set
the ability to create output data sets containing summary statistics, capability indices, histogram
intervals, parameters of fitted curves, and statistical intervals
You can use the PROC CAPABILITY statement, together with the VAR and SPEC statements, to compute
summary statistics and process capability indices. See “Getting Started: CAPABILITY Procedure” on
page 207 for introductory examples. In addition, you can use the statements summarized in Table 6.1 to
request plots and specialized analyses:
Table 6.1 Statements for Plots and Specialized Analyses
Statement Result
CDFPLOT Cumulative distribution function plot
COMPHISTOGRAM Comparative histogram
HISTOGRAM Histogram
INSET Inset table on plot
INTERVALS Statistical intervals
OUTPUT Output data set with summary statistics and capability indices
PPPLOT Probability-probability plot
PROBPLOT Probability plot
QQPLOT Quantile-quantile plot
You have three alternatives for producing plots with the CAPABILITY procedure:
ODS Graphics output is produced if ODS Graphics is enabled, for example by specifying the ODS
GRAPHICS ON statement prior to the PROC statement.
Otherwise, traditional graphics are produced by default if SAS/GRAPH is licensed.
Legacy line printer charts are produced when you specify the LINEPRINTER option in the PROC
statement.
See Chapter 4, “SAS/QC Graphics,” for more information about producing these different kinds of graphs.
You can use the INSET statement with any of the plot statements to enhance the plot with an inset table of
summary statistics. The INSET statement is not applicable when you produce line printer plots.
Learning about the CAPABILITY Procedure F 205
Learning about the CAPABILITY Procedure

To learn about the CAPABILITY procedure, first select the appropriate statement Table 6.1. Then refer to the
corresponding “Getting Started” section for introductory examples:
“Getting Started: CDFPLOT Statement” on page 267
“Getting Started: COMPHISTOGRAM Statement” on page 286
“Getting Started: HISTOGRAM Statement” on page 312
“Getting Started: INSET Statement” on page 399
“Getting Started: INTERVALS Statement” on page 427
“Getting Started: OUTPUT Statement” on page 439
“Getting Started: PPPLOT Statement” on page 455
“Getting Started: PROBPLOT Statement” on page 478
“Getting Started: QQPLOT Statement” on page 509
To broaden your knowledge of the procedure, read “PROC CAPABILITY and General Statements” on
page 205 which summarizes the syntax for the entire procedure and describes the PROC CAPABILITY
statement, the VAR statement, the CLASS statement, and the SPEC statement. Subsequent chapters describe
the statements listed in Table 6.1. In addition to introductory examples, each chapter provides syntax
summaries, descriptions of options, computational details, and advanced examples. Although the chapters
are self-contained, much of what you learn about one plot statement, including the syntax, is transferable to
other plot statements.
PROC CAPABILITY and General Statements
Overview: CAPABILITY Procedure

This chapter describes several statements that are generally used with the CAPABILITY procedure:
The PROC CAPABILITY statement is required to invoke the CAPABILITY procedure. You can use
this statement by itself to compute summary statistics.
The VAR statement, which is optional, specifies the variables in the input data set that are to be
analyzed. These are called the analysis or process variables. By default, all of the numeric variables
are analyzed.
The CLASS statement, which is optional, specifies one or two variables that group the data into
classification levels. A separate analysis is carried out for each combination of levels, and you can use
the CLASS statement with plot statements (such as HISTOGRAM) to create comparative displays. 1
The SPEC statement, which is optional, provides specification limits for the variables that are to be
analyzed. When you use a SPEC statement, the procedure computes process capability indices in
addition to summary statistics. Furthermore, the specification limits are displayed in plots created with
plot statements that are described in subsequent chapters.
You can use the PROC CAPABILITY statement to request a variety of statistics for summarizing the data
distribution of each analysis variable:
sample moments
basic measures of location and variability
confidence intervals for the mean, standard deviation, and variance
tests for location
tests for normality
trimmed and Winsorized means
robust estimates of scale
quantiles and related confidence intervals
extreme observations and extreme values
frequency counts for observations
missing values
You can use the PROC CAPABILITY and SPEC statements together to request a variety of statistics for
process capability analysis:
percents of measurements within and outside specification limits
confidence intervals for the probabilities of exceeding the specification limits
standard capability indices and related confidence intervals
tests of normality in conjunction with capability indices
specialized capability indices
In addition, you can use options in the PROC CAPABILITY statement to
specify the input data set to be analyzed

1 You can use the COMPHISTOGRAM statement to create comparative histograms without applying classification levels to the
overall analysis.
Getting Started: CAPABILITY Procedure F 207
specify an input data set containing specification limits
specify a graphics catalog for saving traditional graphics output
specify rounding units for variable values
specify the definition used to calculate percentiles
specify the divisor used to calculate variances and standard deviations
request legacy line printer plots and define special printing characters used for features
suppress tables
You can use options in the SPEC statement to
provide lower and upper specification limits and target values
control the appearance of specification lines on plots
control the appearance of the areas under a histogram outside the specification limits
Getting Started: CAPABILITY Procedure

This section introduces the PROC CAPABILITY, VAR, and SPEC statements with examples that illustrate
the most commonly used options.
Computing Descriptive Statistics

N OTE : See Computing Summary Stats and Capability Indices in the SAS/QC Sample Library.
The fluid weights of 100 drink cans are measured in ounces. The filling process is assumed to be in statistical
control. The measurements are saved in a SAS data set named Cans.
data Cans;
label Weight = "Fluid Weight (ounces)";
input Weight @@;
datalines;
12.07 12.02 12.00 12.01 11.98 11.96 12.04 12.05 12.01 11.97
12.03 12.03 12.00 12.04 11.96 12.02 12.06 12.00 12.02 11.91
12.05 11.98 11.91 12.01 12.06 12.02 12.05 11.90 12.07 11.98
12.02 12.11 12.00 11.99 11.95 11.98 12.05 12.00 12.10 12.04
12.06 12.04 11.99 12.06 11.99 12.07 11.96 11.97 12.00 11.97
12.09 11.99 11.95 11.99 11.99 11.96 11.94 12.03 12.09 12.03
11.99 12.00 12.05 12.04 12.05 12.01 11.97 11.93 12.00 11.97
12.13 12.07 12.00 11.96 11.99 11.97 12.05 11.94 11.99 12.02
11.95 11.99 11.91 12.06 12.03 12.06 12.05 12.04 12.03 11.98
12.05 12.05 12.11 11.96 12.00 11.96 11.96 12.00 12.01 11.98
;
You can use the PROC CAPABILITY and VAR statements to compute summary statistics for the weights.
title 'Process Capability Analysis of Fluid Weight';

proc capability data=Cans normaltest;
var Weight;
run;
The input data set is specified with the DATA= option. The NORMALTEST option requests tests for
normality. The VAR statement specifies the variables to analyze. If you omit the VAR statement, all numeric
variables in the input data set are analyzed.
The descriptive statistics for Weight are shown in Figure 6.1. For instance, the average weight (labeled Mean)
is 12.0093. The Shapiro-Wilk test statistic labeled W is 0.987876, and the probability of a more extreme
value of W (labeled Pr < W) is 0.499. Compared to the usual cutoff value of 0.05, this probability (referred
to as a p-value) indicates that the weights are normally distributed.
Figure 6.1 Descriptive Statistics

Process Capability Analysis of Fluid Weight

Variable: Weight (Fluid Weight (ounces))
Moments
N 100 Sum Weights 100
Mean 12.0093 Sum Observations 1200.93
Std Deviation 0.04695269 Variance 0.00220456
Skewness 0.05928405 Kurtosis -0.1717404
Uncorrected SS 14422.5469 Corrected SS 0.218251
Coeff Variation 0.39096946 Std Error Mean 0.00469527
Basic Statistical Measures

Location Variability
Mean 12.00930 Std Deviation 0.04695
Median 12.00000 Variance 0.00220
Mode 12.00000 Range 0.23000
Interquartile Range 0.07000
Tests for Location: Mu0=0

Test Statistic p Value
Student's t t 2557.745 Pr > |t| <.0001
Sign M 50 Pr >= |M| <.0001
Signed Rank S 2525 Pr >= |S| <.0001
Tests for Normality

Shapiro-Wilk W 0.987876 Pr < W 0.4991
Kolmogorov-Smirnov D 0.088506 Pr > D 0.0522
Cramer-von Mises W-Sq 0.079055 Pr > W-Sq 0.2179
Anderson-Darling A-Sq 0.457672 Pr > A-Sq >0.2500
Getting Started: CAPABILITY Procedure F 209
Figure 6.1 continued
Quantiles (Definition 5)
Level Quantile
100% Max 12.130
99% 12.120
95% 12.090
90% 12.065
75% Q3 12.050
50% Median 12.000
25% Q1 11.980
10% 11.955
5% 11.935
1% 11.905
0% Min 11.900
Extreme Observations
Lowest Highest
Value Obs Value Obs
11.90 28 12.09 59
11.91 83 12.10 39
11.91 23 12.11 32
11.91 20 12.11 93
11.93 68 12.13 71
Computing Capability Indices

N OTE : See Computing Summary Stats and Capability Indices in the SAS/QC Sample Library.
This example is a continuation of the previous example and shows how you can provide specification limits
with a SPEC statement to request capability indices in addition to descriptive statistics.
proc capability data=Cans normaltest freq;

spec lsl=11.95 target=12 usl=12.05;
var Weight;
run;
The options LSL=, TARGET=, and USL= specify the lower specification limit, target value, and upper
specification limit for the weights. These statements produce the output shown in Figure 6.2 in addition to
the output shown in Figure 6.1.
Figure 6.2 Capability Indices and Frequency Table


Variable: Weight (Fluid Weight (ounces))
Specification Limits
Limit Percent
Lower (LSL) 11.95000 % < LSL 7.00000
Target 12.00000 % Between 77.00000
Upper (USL) 12.05000 % > USL 16.00000
Process Capability Indices

Index Value 95% Confidence Limits
Cp 0.354967 0.305565 0.404288
CPL 0.420991 0.332644 0.508117
CPU 0.288943 0.211699 0.365112
Cpk 0.288943 0.212210 0.365677
Cpm 0.348203 0.301472 0.398228
Frequency Counts
Percents
Value Count Cell Cum
11.90 1 1.0 1.0
11.91 3 3.0 4.0
11.93 1 1.0 5.0
11.94 2 2.0 7.0
11.95 3 3.0 10.0
11.96 8 8.0 18.0
11.97 6 6.0 24.0
11.98 6 6.0 30.0
11.99 10 10.0 40.0
12.00 11 11.0 51.0
12.01 5 5.0 56.0
12.02 6 6.0 62.0
12.03 6 6.0 68.0
12.04 6 6.0 74.0
12.05 10 10.0 84.0
12.06 6 6.0 90.0
12.07 4 4.0 94.0
12.09 2 2.0 96.0
12.10 1 1.0 97.0
12.11 2 2.0 99.0
12.13 1 1.0 100.0
In Figure 6.2, the table labeled Specification Limits lists the specification limits and target value, together with
the percents of observations outside and between the limits. The table labeled Process Capability Indices
lists estimates for the standard process capability indices Cp , CPL, CPU , Cpk , and Cpm , along with 95%
confidence limits. The index Cpm is not computed unless you specify a TARGET= value. See “Standard
Syntax: CAPABILITY Procedure F 211
Capability Indices” on page 246 for formulas used to compute the indices.
If you specify more than one variable in the VAR statement, you can provide corresponding specification
limits and target values by specifying lists of values for the LSL=, USL=, and TARGET= options. As an
alternative to the SPEC statement, you can read specification limits and target values from a data set specified
with the SPEC= option in the PROC CAPABILITY statement. This is illustrated in Example 6.1.
The FREQ option in the PROC CAPABILITY statement requests the table labeled Frequency Counts in
Figure 6.2.
Syntax: CAPABILITY Procedure

The following are the primary statements that control the CAPABILITY procedure:
PROC CAPABILITY < options > ;
BY variables ;
CDFPLOT < variables > < / options > ;
CLASS variable-1 < variable-2 >< / options > ;
COMPHISTOGRAM < variables > / CLASS= (class-variables) < options > ;
FREQ variable ;
HISTOGRAM < variables > < / options > ;
ID variables ;
INSET keyword-list < / options > ;
INTERVALS < variables > < / options > ;
OUTPUT < OUT= SAS-data-set > < keyword1=names . . . keywordk=names > ;
PPPLOT < variables > < / options > ;
PROBPLOT < variables > < / options > ;
QQPLOT < variables > < / options > ;
SPEC < options > ;
VAR variables ;
WEIGHT variable ;
The PROC CAPABILITY statement invokes the procedure. The VAR statement specifies the numeric
variables to be analyzed, and it is required if the OUTPUT statement is used to save summary statistics and
capability indices in an output data set. If you do not use the VAR statement, all numeric variables in the data
set are analyzed. The SPEC statement provides specification limits.
The plot statements (CDFPLOT, COMPHISTOGRAM, HISTOGRAM, PPPLOT, PROBPLOT, and QQ-
PLOT) create graphical displays, and the INSET statement enhances these displays by adding a table of
summary statistics directly on the graph. The INTERVALS statement computes statistical intervals. You can
specify one or more of each of the plot statements, the INSET statement, the INTERVALS statement, and the
OUTPUT statement. If you use a VAR statement, the variables listed in a plot statement must be a subset of
the variables listed in the VAR statement.
PROC CAPABILITY Statement

The syntax for the PROC CAPABILITY statement is as follows:
PROC CAPABILITY < options > ;
The following section lists all options. See the section “Dictionary of Options” on page 214 for detailed
information.
Summary of Options
Table 6.2 lists all the PROC CAPABILITY options by function.
Table 6.2 PROC CAPABILITY Statement Options
Option Description
Input Data Set Options
ANNOTATE= Specifies input data set containing annotation information
DATA= Specifies input data set
EXCLNPWGT Specifies that non-positive weights are to be excluded
NOBYSPECS Specifies that specification limits in SPEC= data set are to be
applied to all BY groups
SPEC= Specifies input data set with specification limits
Plotting and Graphics Options

FORMCHAR(index)= Defines characters used for features on legacy line printer
plots
GOUT= Specifies catalog for saving traditional graphics output
LINEPRINTER Requests legacy line printer plots
Computational Options
FORCEQN Forces calculation of the robust estimator of scale Qn
FORCESN Forces calculation of the robust estimator of scale Sn
PCTLDEF= Specifies definition used to calculate percentiles
ROUND= Specifies units used to round variable values
VARDEF= Specifies divisor used to calculate variances and standard
deviations
Data Summary Options

ALL Requests all tables
FREQ Requests frequency table
MODES Requests table of modes
NEXTROBS= Requests table of n lowest, n highest observations
NEXTRVAL= Requests table of n lowest, n highest values
Output Options
NOPRINT Suppresses printed output
OUTTABLE= Creates an output data set containing univariate statistics and
capability indices in tabular form
Table 6.2 continued
Option Description
Hypothesis Testing Options
MU0= Specifies mean for null hypothesis in tests for location
LOCCOUNT Requests table of counts used in sign test and signed rank test
NORMALTEST Performs tests for normality
Robust Estimation Options

ROBUSTSCALE Requests table of robust measures of scale
TRIMMED=(trimmed- Requests table of trimmed means
options)
WINSORIZED=(Winsorized- Requests table of Winsorized means
options)
TRIMMED-Options
ALPHA= Specifies confidence level
TYPE= Specifies type of confidence limit
WINSORIZED-Options
Capability Index Options
CPMA= (obsolete) Specifies a for Cpm .a/
CHECKINDICES Requests test of normality in conjunction with standard
indices
SPECIALINDICES Requests table of specialized indices including Boyles’ Cpm ,
Cjkp , Cpmk , Cpm .a/, and Wright’s Cs
CHECKINDICES-Options
ALPHA= Specifies cutoff probability for p-values for test for normality
used in conjunction with process capability indices
TEST= Specifies test for normality (Shapiro-Wilk,
Kolmogorov-Smirnov, Anderson-Darling, Cramér–von Mises,
or no test)
Confidence Limit Options

ALPHA= Specifies level for all confidence limits
CIBASIC Requests confidence limits for the mean, standard deviation,
variance
CIINDICES Specifies level and type of confidence limits for capability
indices
CIPCTLDF Requests distribution-free confidence limits for percentiles
CIPCTLNORMAL Requests confidence limits for percentiles assuming normality
CIPROBEX Requests confidence limits for the probability of exceeding
specifications
Table 6.2 continued
Option Description
CIBASIC-Options
CIINDICES-Options
CIPCTLDF-Options
CIPCTLNORMAL-Options
CIPROBEX-Options
Dictionary of Options
The following entries provide detailed descriptions of the options in the PROC CAPABILITY statement.
ALL
requests all of the tables generated by the FREQ, MODES, NEXTRVAL=5, CIBASIC, CIPCTLDF,
and CIPCTLNORMAL options. If a WEIGHT statement is not used, the ALL option also requests
the tables generated by the LOCCOUNT, NORMALTEST, ROBUSTSCALE, TRIMMED=.25, and
WINSORIZED=.25 options. PROC CAPABILITY uses any values that you specify with the ALPHA=,
MUO=, NEXTRVAL=, CIBASIC, CIPCTLDF, CIPCTLNORMAL, TRIMMED=, or WINSORIZED=
options in conjunction with the ALL option.
ALPHA=value
specifies the default confidence level for all confidence limits computed by the CAPABILITY procedure.
The coverage percent for the confidence limits is .1 value/100. For example, ALPHA=0.10 results
in 90% confidence limits. The default value is 0.05.
Note that specialized ALPHA= options are available for a number of confidence interval options. For
example, you can specify CIBASIC( ALPHA=0.10 ) to request a table of Basic Confidence Limits at
the 90% level. The default values of these options default to the value of the general ALPHA= option.
ANNOTATE=SAS-data-set
ANNO=SAS-data-set
specifies an input data set containing annotate variables as described in SAS/GRAPH documentation.
You can use this data set to add features to traditional graphics. Use this data set only when creating
traditional graphics; it is ignored when the LINEPRINTER option is specified and when ODS Graphics
is in effect. Features provided in this data set are added to every plot produced in the current run of the
procedure.
CHECKINDICES< (TEST = SW | KS | AD | CVM | NONE > < ALPHA=value >)

specifies the test of normality used in conjunction with process capability indices that are displayed in
the Process Capability Indices table. If the p-value for the test is less than the cutoff probability value
specified with the ALPHA= option, a warning is added to the table, as illustrated in Figure 6.3. See
“Tests for Normality” on page 238 for details concerning the test.
proc capability data=Process;

var p2;
specs lsl=10
usl=275;
run;
Figure 6.3 Warning Message Printed with Capability Indices


Variable: P2

Cp 0.541072 0.388938 0.692946
CPL 0.642426 0.417087 0.862984
CPU 0.439718 0.257339 0.617184
Cpk 0.439718 0.259310 0.620126
Warning: Normality is rejected for alpha = 0.05 using the Shapiro-Wilk test
ALPHA=value
specifies the cutoff probability for p-values for a test for normality used in conjunction with
process capability indices. The value must be between zero and 0.5. The default value is 0.05.
TEST = SW | KS | AD | CVM | NONE

specifies the test of normality used in conjunction with process capability indices that are
displayed in the Process Capability Indices table. The tests available are Shapiro-Wilk (SW),
Kolmogorov-Smirnov (KS), Anderson-Darling (AD), and Cramér-von Mises (CVM). The default
test is the Shapiro-Wilk test if the sample size is less than or equal to 2000 and the Kolmogorov-
Smirnov test if the sample size is greater than 2000.
CIBASIC< (< TYPE=keyword > < ALPHA=value >) >

requests confidence limits for the mean, standard deviation, and variance based on the assumption
that the data are normally distributed. With large sample sizes, this assumption is not required for
confidence limits for the mean.
ALPHA=value
specifies the confidence level. The coverage percent for the confidence limits is .1 value/100.
For example, ALPHA=0.10 requests 90% confidence limits. The default value is 0.05.
TYPE=keyword
specifies the type of confidence limit, where keyword is LOWER, UPPER, or TWOSIDED. The
default value is TWOSIDED.
CIINDICES< (< TYPE=keyword > < ALPHA=value >) >

specifies the type and level of the confidence limits for standard capability indices displayed in the
table labeled Process Capability Indices.
ALPHA=value
TYPE=keyword
CIPCTLDF< (< TYPE=keyword > < ALPHA=value >) >

CIQUANTDF< (< TYPE=keyword > < ALPHA=value >) >
requests confidence limits for quantiles computed using a distribution-free method. In other words, no
specific parametric distribution (such as the normal) is assumed for the data. Order statistics are used
to compute the confidence limits as described in Section 5.2 of Hahn and Meeker (1991). This option
is not available if you specify a WEIGHT statement.
ALPHA=value
TYPE=keyword
specifies the type of confidence limit, where keyword is LOWER, UPPER, SYMMETRIC, or
ASYMMETRIC. The default value is SYMMETRIC.
CIPCTLNORMAL< (< TYPE=keyword > < ALPHA=value >) >

CIQUANTNORMAL< (< TYPE=keyword > < ALPHA=value >) >
requests confidence limits for quantiles based on the assumption that the data are normally distributed.
The computational method is described in Section 4.4.1 of Hahn and Meeker (1991) and uses the
noncentral t distribution as given by Odeh and Owen (1980). This option is not available if you specify
a WEIGHT statement.
ALPHA=value
TYPE=keyword
CIPROBEX< (< TYPE=keyword > < ALPHA=value >) >

requests confidence limits for P rŒX LSL and P rŒX USL, where X is the analysis variable,
LSL is the lower specification limit, and USL is the upper specification limit. The computational
method, which assumes that X is normally distributed, is described in Section 4.5 of Hahn and Meeker
(1991) and uses the noncentral t distribution as given by Odeh and Owen (1980). This option is not
available if you specify a WEIGHT statement.
ALPHA=value
TYPE=keyword
CPMA=value
specifies the value of the parameter a for the capability index Cpm .a/. This option has been superseded
by the SPECIALINDICES(CPMA=) option.
DATA=SAS-data-set
specifies the input data set containing the observations to be analyzed. If the DATA= option is omitted,
the procedure uses the most recently created SAS data set.
DEF=index
is an alias for the PCTLDEF= option. See the entry for the PCTLDEF= option.
EXCLNPWGT
excludes observations with non-positive weight values (zero or nonnegative) for the analysis. By
default, PROC CAPABILITY treats observations with negative weights like those with zero weights
and counts them in the total number of observations. This option is applicable only if you specify a
WEIGHT statement.
FORCEQN
forces calculation of the robust estimate of scale Qn . Because this calculation is very computationally
intensive, by default Qn is not computed for a variable that has more than 65,526 nonmissing obser-
vations. On some hosts, Qn cannot be computed at all when there are more than 65,526 nonmissing
observations.
FORCESN
forces calculation of the robust estimate of scale Sn . Because this calculation is computationally
intensive, by default Sn is not computed for a variable that has more than 1 million nonmissing
observations.
FORMCHAR(index)=‘string’
defines characters used for features on legacy line printer plots, where index is a number ranging from
1 to 11, and string is a character or hexadecimal string. This option is ignored unless you specify the
LINEPRINTER option in the PROC CAPABILITY statement.
The index identifies which features are controlled with the string characters, as discussed in the table
that follows. If you specify the FORMCHAR= option omitting the index , the string controls all 11
features.
By default, the form character list specified with the SAS system option FORMCHAR= is used;
otherwise, the default is FORMCHAR=’|---|+|--’. If you print to a PC screen or your device
supports the ASCII symbol set (1 or 2), the following is recommended:
formchar='B3,C4,DA,C2,BF,C3,C5,B4,C0,C1,D9'X
As an example, suppose you want to plot the data values of the empirical cumulative distribution
function with asterisks (*). You can change the appropriate character by using the following:
formchar(2)='*'
Note that the FORMCHAR= option in the PROC CAPABILITY statement enables you to temporarily
override the values of the SAS system option with the same name. The values of the SAS system
option are not altered by using the FORMCHAR= option in PROC CAPABILITY statement.
The features associated with values of index are shown in Table 6.3.
Table 6.3 Features Associated with index Values
Value of
index Description of Character Chart Feature
1 Vertical bar Frame, ecdf line, HREF= lines
2 Horizontal bar Frame, ecdf line, VREF= lines
3 Box character (upper left) Frame, ecdf line, histogram bars
4 Box character (upper middle) Histogram bars, tick marks
(horizontal axis)
5 Box character (upper right) Frame, histogram bars
6 Box character (middle left) Histogram bars
7 Box character (middle middle) Not used
8 Box character (middle right) Histogram bars, tick marks
(vertical axis)
9 Box character (lower left) Frame
Table 6.3 continued
Value of
index Description of Character Chart Feature
10 Box character (lower middle) Histogram bars
11 Box character (lower right) Frame, ecdf line
FREQ
requests a frequency table in the printed output that contains the variable values, frequencies, percent-
ages, and cumulative percentages. See Figure 6.2 for an example.
GOUT=graphics-catalog
specifies a graphics catalog in which to save traditional graphics output. This option is ignored unless
you are producing traditional graphics.
LINEPRINTER
requests that legacy line printer plots be produced by the CDFPLOT, HISTOGRAM, PROBPLOT,
PPPLOT, and QQPLOT statements. The CLASS and COMPHISTOGRAM statements cannot be used
when the LINEPRINTER option is specified.
LOCCOUNT
requests a table with the number of observations greater than, not equal to, and less than the value of
MUO=. PROC CAPABILITY uses these values to construct the sign test and signed rank test. This
option is not available if you specify a WEIGHT statement.
MODES
MODE
requests a table of all possible modes. By default, when the data contains multiple modes, PROC
CAPABILITY displays the lowest mode in the table of basic statistical measures. When all values are
unique, PROC CAPABILITY does not produce a table of modes.
MU0=value(s)
LOCATION=value(s)
specifies the value of the mean or location parameter (o ) in the null hypothesis for the tests summarized
in the table labeled Tests for Location: Mu0=value. If you specify a single value, PROC CAPABILITY
tests the same null hypothesis for all analysis variables. If you specify multiple values, a VAR statement
is required, and PROC CAPABILITY tests a different null hypothesis for each analysis variable by
matching the VAR variables with the values in the corresponding order. The default value is 0.
NEXTROBS=n
specifies the number of extreme observations in the table labeled Extreme Observations. The table
lists the n lowest observations and the n highest observations. The default value is 5. The value of n
must be an integer between 0 and half the number of observations. You can specify NEXTROBS=0 to
suppress the table.
NEXTRVAL=n
requests the table labeled Extreme Values and specifies the number of extreme values in the table. The
table lists the n lowest unique values and the n highest unique values. The value of n must be an integer
between 0 and half the maximum number of observations. By default, n = 0 and no table is displayed.
NOBYSPECS
specifies that specification limits in SPEC= data set be applied to all BY groups. If you use a BY
statement and specify a SPEC= data set that does not contain the BY variables, you must specify the
NOBYSPECS option.
NOPRINT
suppresses the tables of descriptive statistics and capability indices which are created by the PROC CA-
PABILITY statement. The NOPRINT option does not suppress the tables created by the INTERVALS
or plot statements. You can use the NOPRINT options in these statements to suppress the creation of
their tables.
NORMALTEST
NORMAL
requests a table of Tests for Normality for each of the analysis variables. The table provides test
statistics and p-values for the Shapiro-Wilk test (provided the sample size is less than or equal to 2000),
the Kolmogorov-Smirnov test, the Anderson-Darling test, and the Cramér–von Mises test. See “Tests
for Normality” on page 238 for details. If specification limits are provided, the NORMALTEST option
is assumed.
OUTTABLE=SAS-data-set
specifies an output data set that contains univariate statistics and capability indices arranged in tabular
form. See “OUTTABLE= Data Set” on page 232 for details.
PCTLDEF=index
DEF=index
specifies one of five definitions used to calculate percentiles. The value of index can be 1, 2, 3, 4, or 5.
See “Percentile Computations” on page 240 for details. By default, PCTLDEF=5.
ROBUSTSCALE
requests a table of robust measures of scale. These measures include the interquartile range, Gini’s
mean difference, the median absolute deviation about the median (MAD), and two statistics proposed
by Rousseeuw and Croux (1993), Qn , and Sn . This option is not available if you specify a WEIGHT
statement.
ROUND=value-list
specifies units used to round variable values. The ROUND= option reduces the number of unique
values for each variable and hence reduces the memory required for temporary storage. Values must be
greater than 0 for rounding to occur.
If you use only one value, the procedure uses this unit for all variables. If you use a list of values, you
must also use a VAR statement. The procedure then uses the roundoff values for variables in the order
given in the VAR statement. For example, the following statements specify a roundoff value of 1 for
Yieldstrength and a roundoff value of 0.5 for TENSTREN.
proc capability round=1 0.5;

var Yieldstrength tenstren;
run;
When a variable value is midway between the two nearest rounded points, the value is rounded to the
nearest even multiple of the roundoff value. For example, with a roundoff value of 1, the variable
values of –2.5, –2.2, and –1.5 are rounded to –2; the values of –0.5, 0.2, and 0.5 are rounded to 0; and
the values of 0.6, 1.2, and 1.4 are rounded to 1.
SPECIALINDICES
requests a table of specialized process capability indices. These indices include k, Boyles’ modified
Cpm (also denoted as Cpm C), Cjkp , Cpm .a/, Cp.5:15/ , Cpk .5:15/ , Cpmk , Wright’s Cs , Boyles’ Sjkp ,
00 W
Cpp , Cpp , Cpg , Cpq , CpW , Cpk W , C , and Vännmann’s C .u; v/ and C .v/.
, Cpm pc p p
You can provide values for the parameters a for Cpm .a/, u and v for Cp .u; v/ and Cp .v/, and for the
multiplier for Cs by specifying the following options in parentheses after the SPECIALINDICES
option.
CPMA=value
specifies the value of the parameter a for the capability index Cpm .a/ described in Section 3.7
of Kotz and Johnson (1993). The value must be positive. The default value is 0.5. The existing
CPMA= option in the PROC CAPABILITY statement is considered obsolete but still works.
CPU=value
specifies the value of the parameter u for Vännmann’s capability index Cp .u; v/. The value must
be greater than or equal to zero. The default value is zero.
CPV=value
specifies the value of the parameter v for Vännmann’s capability indices Cp .u; v/ and Cp .v/.
The value must be greater than or equal to zero. The default value is 4.
CSGAMMA=value
specifies the value of the multiplier suggested by Chen and Kotz (1996) for Wright’s capability
index Cs . The value must be greater than zero. The default value is 1.
SPEC=SAS-data-set
SPECS=SAS-data-set
specifies an input data set containing specification limits for each of the variables in the VAR statement.
This option is an alternative to the SPEC statement, which also provides specification limits. See
“SPEC= Data Set” on page 230 for details on SPEC= data sets, and Example 6.1 for an example. If you
use both the SPEC= option and a SPEC statement, the SPEC= option is ignored.
TRIMMED=values(s) < (TYPE=keyword >< ALPHA=value >)

requests a table of trimmed means, where each value specifies the number or the proportion of trimmed
observations. If the value is the number n of trimmed observations, n must be between 0 and half the
number of nonmissing observations. If the value is a proportion p between 0 and 0.5, the number of
observations trimmed is the smallest integer greater than or equal to np, where n is the number of
observations. To obtain confidence limits for the mean and the student t-test, you must use the default
value of VARDEF= which is DF. The TRIMMED= option is not available if you specify a WEIGHT
statement.
ALPHA=value
specifies the confidence level. The coverage percent is .1 value/100. For example, AL-
PHA=0.10 requests a 90% confidence limit. The default value is 0.05.
TYPE=keyword
VARDEF=DF | N | WDF | WEIGHT | WGT

specifies the divisor used in calculating variances and standard deviations. The values and associated
divisors are shown in the following table. By default, VARDEF=DF.
Value Divisor Formula

DF Degrees of freedom n 1
N Number of observations n
P
WEIGHT | WGT Sum of weight i wi
P
WDF Sum of weights minus one . i wi / 1
WINSORIZED=values(s) < (< TYPE=keyword > < ALPHA=value >) >

WINSOR=values(s) < (< TYPE=keyword > < ALPHA=value >) >
requests a table of winsorized means, where each value specifies the number or the proportion of
winsorized observations. If the value is the number n of winsorized observations, n must be between
0 and half the number of nonmissing observations. If the value is a proportion p between 0 and 0.5,
the number of observations winsorized is the smallest integer greater than or equal to np, where n is
the number of observations. To obtain confidence limits for the mean and the student t-test, you must
use the default value of VARDEF= which is DF. The WINSORIZED= option is not available if you
specify a WEIGHT statement.
ALPHA=value
specifies the confidence level. The coverage percent is .1 value/100. For example, AL-
PHA=0.10 results in a 90% confidence limit. The default value is 0.05.
TYPE=keyword
BY Statement
BY variables ;
You can specify a BY statement in PROC CAPABILITY to obtain separate analyses of observations in groups
that are defined by the BY variables. When a BY statement appears, the procedure expects the input data
set to be sorted in order of the BY variables. If you specify more than one BY statement, only the last one
specified is used.
If your input data set is not sorted in ascending order, use one of the following alternatives:
Sort the data by using the SORT procedure with a similar BY statement.
Specify the NOTSORTED or DESCENDING option in the BY statement in the CAPABILITY

procedure. The NOTSORTED option does not mean that the data are unsorted but rather that the
data are arranged in groups (according to values of the BY variables) and that these groups are not
necessarily in alphabetical or increasing numeric order.
Create an index on the BY variables by using the DATASETS procedure (in Base SAS software).
For more information about BY-group processing, see the discussion in SAS Language Reference: Concepts.
For more information about the DATASETS procedure, see the discussion in the Base SAS Procedures Guide.
CLASS Statement
CLASS variable-1 < (v-options) > < variable-2 < (v-options) > >
< / KEYLEVEL= value1 | (value1 value2) > ;
The CLASS statement specifies one or two variables used to group the data into classification levels. Variables
in a CLASS statement are referred to as CLASS variables. CLASS variables can be numeric or character.
Class variables can have floating point values, but they typically have a few discrete values that define
levels of the variable. You do not have to sort the data by CLASS variables. PROC CAPABILITY uses the
formatted values of the CLASS variables to determine the classification levels.
N OTE : You cannot specify a COMPHISTOGRAM statement together with a CLASS statement.
You can specify the following v-options enclosed in parentheses after a CLASS variable:
MISSING
specifies that missing values for the CLASS variable are to be treated as valid classification levels.
Special missing values that represent numeric values (‘.A’ through ‘.Z’ and ‘._’) are each considered
as a separate value. If you omit MISSING, PROC CAPABILITY excludes the observations with a
missing CLASS variable value from the analysis. Enclose this option in parentheses after the CLASS
variable.
ORDER=DATA | FORMATTED | FREQ | INTERNAL

specifies the display order for the CLASS variable values. The default value is INTERNAL. You can
specify the following values with the ORDER= option:
DATA orders values according to their order in the input data set. When you use a plot
statement, PROC CAPABILITY displays the rows (columns) of the comparative
plot from top to bottom (left to right) in the order that the CLASS variable values
first appear in the input data set.
FORMATTED orders values by their ascending formatted values. This order might depend on
your operating environment. When you use a plot statement, PROC CAPABILITY
displays the rows (columns) of the comparative plot from top to bottom (left to
right) in increasing order of the formatted CLASS variable values. For example,
suppose a numeric CLASS variable DAY (with values 1, 2, and 3) has a user-defined
format that assigns Wednesday to the value 1, Thursday to the value 2, and Friday
to the value 3. The rows of the comparative plot will appear in alphabetical order
(Friday, Thursday, Wednesday) from top to bottom.
If there are two or more distinct internal values with the same formatted value, then
PROC CAPABILITY determines the order by the internal value that occurs first in
the input data set. For numeric variables without an explicit format, the levels are
ordered by their internal values.
FREQ orders values by descending frequency count so that levels with the most observa-
tions are listed first. If two or more values have the same frequency count, PROC
CAPABILITY uses the formatted values to determine the order.
When you use a plot statement, PROC CAPABILITY displays the rows (columns)
of the comparative plot from top to bottom (left to right) in order of decreasing
frequency count for the CLASS variable values.
INTERNAL orders values by their unformatted values, which yields the same order as PROC
SORT. This order may depend on your operating environment.
When you use a plot statement, PROC CAPABILITY displays the rows (columns)
of the comparative plot from top to bottom (left to right) in increasing order of the
internal (unformatted) values of the CLASS variable. The first CLASS variable
is used to label the rows of the comparative plots (top to bottom). The second
CLASS variable is used to label the columns of the comparative plots (left to right).
For example, suppose a numeric CLASS variable DAY (with values 1, 2, and 3)
has a user-defined format that assigns Wednesday to the value 1, Thursday to the
value 2, and Friday to the value 3. The rows of the comparative plot will appear in
day-of-the-week order (Wednesday, Thursday, Friday) from top to bottom.
You can specify the following options after the slash (/) in the CLASS statement.
KEYLEVEL=value | ( value1 value2 )

specifies the key cells in comparative plots. For each plot, PROC CAPABILITY first determines the
horizontal axis scaling for the key cell, and then extends the axis using the established tick interval to
accommodate the data ranges for the remaining cells, if necessary. Thus, the choice of the key cell
determines the uniform horizontal axis that PROC CAPABILITY uses for all cells.
If you specify only one CLASS variable and use a plot statement, KEYLEVEL=value identifies the
key cell as the level for which the CLASS variable is equal to value. By default, PROC CAPABILITY
sorts the levels in the order determined by the ORDER= option, and the key cell is the first occurrence
of a level in this order. The cells display in order from top to bottom or left to right. Consequently,
the key cell appears at the top (or left). When you specify a different key cell with the KEYLEVEL=
option, this cell appears at the top (or left).
If you specify two CLASS variables, use KEYLEVEL= ( value1 value2) to identify the key cell as the
level for which CLASS variable n is equal to valuen. By default, PROC CAPABILITY sorts the levels
of the first CLASS variable in the order that is determined by its ORDER= option. Then, within each
of these levels, it sorts the levels of the second CLASS variable in the order that is determined by its
ORDER= option. The default key cell is the first occurrence of a combination of levels for the two
variables in this order. The cells display in the order of the first CLASS variable from top to bottom
and in the order of the second CLASS variable from left to right. Consequently, the default key cell
appears at the upper left corner. When you specify a different key cell with the KEYLEVEL= option,
this cell appears at the upper left corner.
The length of the KEYLEVEL= value cannot exceed 16 characters and you must specify a formatted
value.
The KEYLEVEL= option has no effect unless you specify a plot statement.
NOKEYMOVE
specifies that the location of the key cell in a comparative plot be unchanged by the CLASS statement
KEYLEVEL= option. By default, the key cell is positioned as the first cell in a comparative plot.
The NOKEYMOVE option has no effect unless you specify a plot statement.
FREQ Statement
FREQ variable ;
The FREQ statement names a variable that provides frequencies for each observation in the input data set. If
n is the value of the FREQ variable for a particular observation, then that observation is used n times. If the
value of the FREQ variable is missing or is less than one, the observation is not used in the analysis. If the
value is not an integer, only the integer portion is used.
ID Statement
ID variables ;
The ID statement specifies one or more variables to include in the table of extreme observations. The
corresponding values of the ID variables appear beside the n largest and n smallest observations, where n is
the value of the NEXTROBS= option.
SPEC Statement
The syntax for the SPEC statement is as follows:
SPEC < options > ;
You can use at most one SPEC statement in the CAPABILITY procedure. When you provide specification
limits and target values in a SPEC statement, the tabular output produced by the PROC CAPABILITY
statement includes process capability indices as well as summary statistics. You can use the SPEC statement
in conjunction with the CDFPLOT, COMPHISTOGRAM, HISTOGRAM, PROBPLOT, and QQPLOT
statements to add specification limit and target lines to the plots produced with these statements.
options
control features of the specification limits and target values. Table 6.4 lists all options by function.
Summary of Options
Table 6.4 SPEC Statement Options
Option Description
Lower Specification Limit Options
CLEFT= Color used to fill area left of lower specification limit (histograms
only)
CLSL= Color of lower specification limit line
LLSL= Line type of lower specification limit line
LSL= Lower specification limit values
LSLSYMBOL= Character used for lower specification limit line in line printer plots
Table 6.4 continued
Option Description
PLEFT= Pattern type used to fill area left of lower specification limit
(histograms only)
WLSL= Width of lower specification limit line
Target Options
CTARGET= Color of target line
LTARGET= Line type of target line
TARGET= Target value
TARGETSYMBOL= Character used for target in line printer plots
WTARGET= Width of target line
Upper Specification Limit Options

CRIGHT= Color used to fill area right of upper specification limit (histograms
only)
CUSL= Color of upper specification limit line
LUSL= Line type of upper specification limit line
PRIGHT= Pattern type used to fill area right of upper specification limit
(histograms only)
USL= Upper specification limit values
USLSYMBOL= Character used for upper specification limit in line printer plots
WUSL= Width of upper specification limit line
General Options
You can specify the following options whether you are producing ODS Graphics output or traditional
graphics:
CLEFT=color
CLEFT
determines the color used to fill the area under a histogram to the left of the lower specification limit.
You can specify the CLEFT option without an argument to fill this area with an appropriate color from
the ODS style. If you are producing ODS Graphics output, an explicit color specification is ignored.
This option is applicable only when the SPEC statement is used in conjunction with a HISTOGRAM
or COMPHISTOGRAM statement. See Output 6.2.1 for an example. The CLEFT= option also applies
to the area under a fitted curve;for an example, see Output 6.8.1.
CRIGHT=color
CRIGHT
determines the color used to fill the area under a histogram to the right of the upper specification
limit. You can specify the CRIGHT option without an argument to fill this area with an appropriate
color from the ODS style. If you are producing ODS Graphics output, an explicit color specification
is ignored. This option is applicable only when the SPEC statement is used in conjunction with a
HISTOGRAM or COMPHISTOGRAM statement. See Output 6.2.1 for an example. The CRIGHT=
option also applies to the area under a fitted curve; for an example, see Output 6.8.1.
LSL=value-list
specifies the lower specification limits for the variables listed in the VAR statement, or for all numeric
variables in the input data set if no VAR statement is used. If you specify only one lower limit, it is
used for all of the variables; otherwise, the number of limits must match the number of variables. See
the section “Computing Capability Indices” on page 209 for an example.
TARGET=value-list
specifies target values for the variables listed in the VAR statement, or for all numeric variables in the
input data set if no VAR statement is used. If you specify only one target value, it is used for all of
the variables; otherwise, the number of values must match the number of variables. See the section
“Computing Capability Indices” on page 209 for an example.
USL=value-list
specifies the upper specification limits for the variables listed in the VAR statement, or for all numeric
variables in the input data set if no VAR statement is used. If you specify only one upper limit, it is
used for all of the variables; otherwise, the number of limits must match the number of variables. See
the section “Computing Capability Indices” on page 209 for an example.
Options for Traditional Graphics

You can specify the following options if you are producing traditional graphics:
CLSL=color
specifies the color of the lower specification line displayed in plots created with the CDFPLOT,
COMPHISTOGRAM, HISTOGRAM, PROBPLOT, and QQPLOT statements.
CTARGET=color
specifies the color of the target line displayed in plots created with the CDFPLOT, COMPHIS-
TOGRAM, HISTOGRAM, PROBPLOT, and QQPLOT statements.
CUSL=color
specifies the color of the upper specification line displayed in plots created with the CDFPLOT,
COMPHISTOGRAM, HISTOGRAM, PROBPLOT, and QQPLOT statements.
LLSL=linetype
specifies the line type for the lower specification line displayed in plots created with the CDFPLOT,
COMPHISTOGRAM, HISTOGRAM, PROBPLOT, and QQPLOT statements. See Output 6.2.1 for an
example. The default is 1, which produces a solid line.
LTARGET=linetype
specifies the line type for the target line in plots created with the CDFPLOT, COMPHISTOGRAM,
HISTOGRAM, PROBPLOT, and QQPLOT statements. See Output 6.2.1 for an example. The default
is 1, which produces a solid line.
LUSL=linetype
specifies the line type for the upper specification line displayed in plots created with the CDFPLOT,
example. The default is 1, which produces a solid line.
PLEFT=pattern
specifies the pattern used to fill the area under a histogram to the left of the lower specification limit.
This option is applicable only when the SPEC statement is used in conjunction with a HISTOGRAM or
COMPHISTOGRAM statement. For an example, see Output 6.2.1. The PLEFT= option also applies
to the area under a fitted curve; for an example, see Output 6.8.1. The default pattern is a solid fill.
PRIGHT=pattern
specifies the pattern used to fill the area under a histogram to the right of the upper specification limit.
This option is applicable only when the SPEC statement is used in conjunction with a HISTOGRAM or
COMPHISTOGRAM statement. For an example, see Output 6.2.1. The PRIGHT= option also applies
to the area under a fitted curve; for an example, see Output 6.8.1. The default pattern is a solid fill.
WLSL=n
specifies the width in pixels of the lower specification line in plots created with the CDFPLOT,
illustration. The default is 1.
WTARGET=n
specifies the width in pixels of the target line in plots created with the CDFPLOT, COMPHISTOGRAM,
HISTOGRAM, PROBPLOT, and QQPLOT statements. See Output 6.2.1 for an illustration. The
default is 1.
WUSL=n
specifies the width in pixels of the upper specification line in plots created with the CDFPLOT,
illustration. The default is 1.
Options for Legacy Line Printer Plots

You can specify the following options if you are producing legacy line printer plots:
LSLSYMBOL=‘character ’
specifies the character used to display the lower specification line in line printer plots created with the
CDFPLOT, HISTOGRAM, PROBPLOT, and QQPLOT statements. The default character is ‘L’.
TARGETSYMBOL=‘character ’
TARGETSYM=‘character’
specifies the character used to display the target line in line printer plots created with the CDFPLOT,
HISTOGRAM, PROBPLOT, and QQPLOT statements. The default character is ‘T’.
USLSYMBOL=‘character ’
specifies the character used to display the upper specification line in line printer plots created with the
CDFPLOT, HISTOGRAM, PROBPLOT, and QQPLOT statements. The default character is ‘U’.
VAR Statement
VAR variables ;
The VAR statement specifies the analysis variables and their order in the results. By default, if you omit
the VAR statement, PROC CAPABILITY analyzes all numeric variables that are not listed in the other
statements.
You must provide a VAR statement when you use an OUTPUT statement. To store the same statistic for
several analysis variables in the OUT= data set, you specify a list of names in the OUTPUT statement. PROC
CAPABILITY makes a one-to-one correspondence between the order of the analysis variables in the VAR
statement and the list of names that follow a statistic keyword.
WEIGHT Statement
WEIGHT variable ;
The WEIGHT statement names a variable that provides weights for each observation in the input data set. The
CAPABILITY procedure uses the values wi of the WEIGHT variable to modify the computation of a number
of summary statistics by assuming that the variance of the ith value Xi of the analysis variable is equal to
2 =wi , where is an unknown parameter. This assumption is rarely applicable in process capability analysis,
and the purpose of the WEIGHT statement is simply to make the CAPABILITY procedure consistent with
other data summarization procedures, such as the UNIVARIATE procedure.
The values of the WEIGHT variable do not have to be integers and are typically positive. By default,
observations with non-positive or missing values of the WEIGHT variable are handled as follows:
If the value is zero, the observation is counted in the total number of observations.
If the value is negative, it is converted to zero, and the observation is counted in the total number of
observations.
If the value is missing, the observation is excluded from the analysis.
To exclude observations that contain negative and zero weights from the analysis, specify the option
EXCLNPWGT in the PROC statement. Note that most SAS/STAT procedures, such as PROC GLM, exclude
negative and zero weights by default.
When you specify a WEIGHT variable, the procedure uses its values, wi , to compute weighted versions of
the statistics provided in the Moments
P table. For example, the procedure computes a weighted mean X w and
2 i wi xi 1 P
a weighted variance sw as X w D P and sw D d i wi .xi X w /2 where xi is the ith variable value.
2
i wi
The divisor d is controlled by the VARDEF= option in the PROC CAPABILITY statement.
When you use both the WEIGHT and SPEC statements, capability indices are computed using X w and sw in
place of X and s. Again, note that weighted capability indices are seldom needed in practice.
When you specify a WEIGHT statement, the procedure also computes a weighted standard error and a
weighted version of Student’s t test. This test is the only test of location that is provided when weights are
specified.
The WEIGHT statement does not affect the determination of the mode, extreme values, extreme observations,
or the number of missing values of the analysis variables. However, the weights wi are used to compute
weighted percentiles.
The WEIGHT variable has no effect on the calculation of extreme values, and it has no effect on graphical
displays produced with the plot statements.
Graphical Enhancement Statements

You can use TITLE, FOOTNOTE, and NOTE statements to enhance printed output. If you are creating
traditional graphics, you can also use AXIS, LEGEND, PATTERN, and SYMBOL statements to enhance
your plots. For details, see SAS/GRAPH documentation and the chapter for the plot statement that you are
using.
Details: CAPABILITY Procedure

This section provides details on the following topics:
input data sets specified with the DATA= option, the SPEC= option, and the ANNOTATE= option
the output data set specified with the OUTTABLE= option
descriptive statistics
the tests for normality requested with the NORMALTEST option
percentile definitions controlled using the PCTLDEF= option
robust estimators
computing the mode
assumptions and terminology for capability indices
standard capability indices
specialized capability indices
Input Data Sets

DATA= Data Set
The DATA= data set contains a set of variables that represent measurements from a process. The CAPABIL-
ITY procedure must have a DATA= data set. If you do not specify one with the DATA= option in the PROC
CAPABILITY statement, the procedure uses the last data set created.
SPEC= Data Set

The SPEC= option in the PROC CAPABILITY statement identifies a SPEC= data set, which contains
specification limits. This option is an alternative to using the SPEC statement. If you use both the SPEC=
option and a SPEC statement, the SPEC= option is ignored. The SPEC= option is especially useful when:
the number of variables is large
the same specification limits are referred to in more than one analysis
Details: CAPABILITY Procedure F 231
a BY statement is used
batch processing is used
The following variables are read from a SPEC= data set:
Variable Description
_LSL_ Lower specification limit
_TARGET_ Target value
_USL_ Upper specification limit
_VAR_ Name of the variable
You may omit either _LSL_ or _USL_ but not both. _TARGET_ is optional. If the SPEC= data set
contains both _LSL_ and _USL_, you can assign missing values to _LSL_ or _USL_ to indicate one-sided
specifications. You can assign missing values to _TARGET_ when the variable does not use a target value.
_LSL_, _USL_, and _TARGET_ must be numeric variables. _VAR_ must be a character variable.
You can include the optional variables listed in Table 6.5 in a SPEC= data set to control the appearance of
specification limits on charts.
Table 6.5 SPEC= Data Set Optional Variables
_CLEFT_ Color used to fill area left of LSL (histograms only)
_CLSL_ Color of LSL line
_CRIGHT_ Color used to fill area right of USL (histograms only)
_CTARGET_ Color of target line
_CUSL_ Color of USL line
_LLSL_ Line type of LSL line
_LSLSYM_ Character used for LSL line in line printer plots
_LTARGET_ Line type of target line
_LUSL_ Line type of USL line
_PLEFT_ Pattern type used to fill area left of LSL (histograms only)
_PRIGHT_ Pattern type used to fill area right of USL (histograms only)
_TARGETSYM_ Character used for target in line printer plots
_USLSYM_ Character used for USL line in line printer plots
_WLSL_ Width of LSL line
_WTARGET_ Width of target line
_WUSL_ Width of USL line
If you are using the HISTOGRAM statement to create “clickable” histograms in HTML, you can also provide
the following variables in a SPEC= data set:
_LOURL_ URL associated with area to left of lower specification limit
_HIURL_ URL associated with area to right of upper specification limit
_URL_ URL associated with area between specification limits
These are character variables whose values are Uniform Resource Locators (URLs) linked to areas on a
histogram. When you view the ODS HTML output with a browser, you can click on an area, and the browser
will bring up the page specified by the corresponding URL.
If you use a BY statement, the SPEC= data set must also contain the BY variables. The SPEC= data set must
be sorted in the same order as the DATA= data set. Within a BY group, specification limits for each variable
plotted are read from the first observation where _VAR_ matches the variable name.
See the section “Examples: CAPABILITY Procedure” on page 260 for an example of reading specification
limits from a SPEC= data set.
ANNOTATE= Data Sets

In the CAPABILITY procedure, you can add features to traditional graphics plots by specifying ANNOTATE=
data sets either in the PROC CAPABILITY statement or in individual plot statements. Depending on where
you specify an ANNOTATE= data set, however, the information is used for all plots or only for plots produced
by a particular statement.
Information contained in the ANNOTATE= data set specified in the PROC CAPABILITY statement is used
for all plots produced in a particular PROC step; this is a “global” ANNOTATE= data set. By using this
global data set, you can keep information common to all plots in one data set.
Information contained in the ANNOTATE= data set specified in a plot statement is used for plots produced by
that statement; this is a “local” ANNOTATE= data set. By using this data set, you can add statement-specific
features to plots. For example, you can add different features to plots produced by the HISTOGRAM and
QQPLOT statements by specifying an ANNOTATE= data set in each plot statement.
In addition, you can specify an ANNOTATE= data set in the PROC CAPABILITY statement and in plot
statements. This enables you to add some features to all plots (those given in the data set specified in the
PROC statement) and also add statement-specific features to plots (those given in the data set specified in the
plot statement).
For complete details on the structure and content of Annotate type data sets, see SAS/GRAPH documentation.
Output Data Set

OUTTABLE= Data Set
The OUTTABLE= data set saves univariate statistics and capability indices. Table 6.6 lists the variables that
can be saved.
Table 6.6 OUTTABLE= Data Set
_CP_ Capability index Cp
_CPLCL_ Lower confidence limit for Cp
_CPUCL_ Upper confidence limit for Cp
_CPK_ Capability index Cpk
_CPKLCL_ Lower confidence limit for Cpk
_CPKUCL_ Upper confidence limit for Cpk
_CPL_ Capability index CPL
_CPLLCL_ Lower confidence limit for CPL
_CPLUCL_ Upper confidence limit for CPL
_CPM_ Capability index Cpm
_CPMLCL_ Lower confidence limit for Cpm
_CPMUCL_ Upper confidence limit for Cpm
_CPU_ Capability index CPU
_CPULCL_ Lower confidence limit for CPU
_CPUUCL_ Upper confidence limit for CPU
_CSS_ Corrected sum of squares
_CV_ Coefficient of variation
_GEOMEAN_ Geometric mean
_GINI_ Gini’s mean difference
_HARMEAN_ Harmonic mean
_K_ Capability index K
_KURT_ Kurtosis
_MAD_ Median absolute difference about the median
_MAX_ Maximum
_MEAN_ Mean
_MEDIAN_ Median
_MIN_ Minimum
_MODE_ Mode
_MSIGN_ Sign statistic
_NMISS_ Number of missing observations
_NOBS_ Number of nonmissing observations
_P1_ 1st percentile
_P5_ 5th percentile
_P10_ 10th percentile
_PCTGTR_ Percentage of observations greater than upper specification limit
_PCTLSS_ Percentage of observations less than lower specification limit
_PROBM_ p-value of sign statistic
_PROBN_ p-value of test for normality
_PROBS_ p-value of signed rank test
_PROBT_ p-value of t statistic
Table 6.6 continued
_Q1_ 25th percentile (lower quartile)
_Q3_ 75th percentile (upper quartile)
_QN_ Qn (see “Robust Estimates of Scale” on page 244)
_QRANGE_ Interquartile range (upper quartile minus lower quartile)
_RANGE_ Range
_SGNRNK_ Centered sign rank
_SKEW_ Skewness
_SN_ Sn (see “Robust Estimates of Scale” on page 244)
_STD_ Standard deviation
_STDGINI_ Gini’s standard deviation
_STDMAD_ MAD standard deviation
_STDMEAN_ Standard error of the mean
_STDQN_ Qn standard deviation
_STDQRANGE_ Interquartile range standard deviation
_STDSN_ Sn standard deviation
_SUMWGT_ Sum of the weights
_SUM_ Sum
_USS_ Uncorrected sum of squares
_VARI_ Variance
_VAR_ Variable name
N OTE : The variables _CP_, _CPLCL_, _CPUCL_, _CPK_, _CPKLCL_, _CPKUCL_, _CPL_, _CPLLCL_,
_CPLUCL_, _CPM_, _CPMLCL_, _CPMUCL_, _CPU_, _CPULCL_, _CPUUCL_, _K_, _LSL_, _PCTGTR_,
_PCTLSS_, _TARGET_, and _USL_ are included if you provide specification limits.
The OUTTABLE= data set and the OUT= data set2 contain essentially the same information. However,
the structure of the OUTTABLE= data set may be more appropriate when you are computing summary
statistics or capability indices for more than one process variable in the same invocation of the CAPABILITY
procedure. Each observation in the OUTTABLE= data set corresponds to a different process variable, and
the variables in the data set correspond to summary statistics and indices.
N OTE : See Tabulating Results for Multiple Variables in the SAS/QC Sample Library.
For example, suppose you have ten process variables (P1-P10). The following statements create an OUT-
TABLE= data set named Table, which contains summary statistics and capability indices for each of these
variables:
proc capability data=Process outtable=Table noprint;

var P1-P10;
specs lsl=5 10 65 35 35 5 25 25 60 15
usl=175 275 300 450 550 200 275 425 500 525;
run;
2 See “OUTPUT Statement: CAPABILITY Procedure” on page 439 for details on the OUT= data set.
The following statements create the table shown in Figure 6.4, which contains the mean, standard deviation,
lower and upper specification limits, and capability index Cpk for each process variable:
proc print data=Table label noobs;

var _VAR_ _MEAN_ _STD_ _LSL_ _USL_ _CPK_;
label _VAR_='Process';
run;
Figure 6.4 Tabulating Results for Multiple Process Variables
Lower Upper Capability

Standard Specification Specification Index
Process Mean Deviation Limit Limit CPK
P1 90.76 57.024 5 175 0.49242
P2 167.32 81.628 10 275 0.43972
P3 224.56 96.525 65 300 0.26052
P4 258.08 145.218 35 450 0.44053
P5 283.48 157.033 35 550 0.52745
P6 107.48 52.437 5 200 0.58814
P7 153.20 90.031 25 275 0.45096
P8 217.08 130.031 25 425 0.49239
P9 280.68 140.943 60 500 0.51870
P10 243.24 178.799 15 525 0.42551
Descriptive Statistics
This section provides computational details for the descriptive statistics which are computed with the PROC
CAPABILITY statement. These statistics can also be saved in the OUT= data set by specifying the keywords
listed in Table 6.59 in the OUTPUT statement.
Standard algorithms (Fisher 1973) are used to compute the moment statistics. The computational methods
used by the CAPABILITY procedure are consistent with those used by other SAS procedures for calculating
descriptive statistics. For details on statistics also calculated by Base SAS software, see Base SAS Procedures
Guide.
The following sections give specific details on several statistics calculated by the CAPABILITY procedure.
Mean
The sample mean is calculated as
Pn
wi xi
PiD1
n
i D1 wi
where n is the number of nonmissing values for a variable, xi is the ith value of the variable, and wi is the
weight associated with the ith value of the variable. If there is no WEIGHT= variable, the formula reduces to
1 Pn
n i D1 xi .
Sum
The sum is calculated as niD1 wi xi , where n is the number of nonmissing values for a variable, xi is the
P
ith value of the variable, and wi is the weight
Pnassociated with the ith value of the variable. If there is no
WEIGHT= variable, the formula reduces to i D1 xi .
Sum of the Weights

Pn
The sum of the weights is calculated as i D1 wi , where n is the number of nonmissing values for a variable
and wi is the weight associated with the ith value of the variable. If there is no WEIGHT= variable, the sum
of the weights is n.
Variance
The variance is calculated as
n
1X
wi .xi XN w /2
d
i D1
where n is the number of nonmissing values for a variable, xi is the ith value of the variable, XN w is the
weighted mean, wi is the weight associated with the ith value of the variable, and d is the divisor controlled
by the VARDEF= option in the PROC CAPABILITY statement. If there is no WEIGHT= variable, the
formula reduces to
n
1X
.xi XN w /2
d
i D1
Standard Deviation
The standard deviation is calculated as
v
u n
u1 X
t wi .xi XN w /2
d
iD1
where n is the number of nonmissing values for a variable, xi is the ith value of the variable, XN w is the
weighted mean, wi is the weight associated with the ith value of the variable, and d is the divisor controlled
by the VARDEF= option in the PROC CAPABILITY statement. If there is no WEIGHT= variable, the
formula reduces to
v
u n
u1 X
t .xi XN w /2
d
iD1
Skewness
The sample skewness is calculated as
n
!3
n X xi XN
.n 1/.n 2/ s
i D1
where n is the number of nonmissing values for a variable and must be greater than 2, xi is the ith value of
the variable, XN is the sample average, and s is the sample standard deviation.
The sample skewness can be positive or negative; it measures the asymmetry of the data distribution and
p 3
estimates the theoretical skewness ˇ1 D 3 2 2 , where 2 and 3 are the second and third central
moments. Observations that are normally distributed should have a skewness near zero.
Kurtosis
The sample kurtosis is calculated as
n
!4
n.n C 1/ X xi XN 3.n 1/2
.n 1/.n 2/.n 3/ s .n 2/.n 3/
i D1
where n > 3. The sample kurtosis measures the heaviness of the tails of the data distribution. It estimates
the adjusted theoretical kurtosis denoted as ˇ2 3, where ˇ2 D 42 , and 4 is the fourth central moment.
2
Observations that are normally distributed should have a kurtosis near zero.
Coefficient of Variation (CV)

The coefficient of variation is calculated as
100 s
CV D
XN
Geometric Mean
The geometric mean is calculated as
n
!1= PniD1 wi
xiwi
Y
i D1
where n is the number of nonmissing values for a variable, xi is the ith value of the variable, and wi is the
weight associated with the ith value of the variable.
If there is no WEIGHT variable, the formula reduces to
n
!1=n
Y
xi
i D1
If any xi is negative, the geometric mean is set to missing.
Signed Rank Statistic

The signed rank statistic S is computed as
X n.n C 1/
SD riC
4
i Wxi >0
where riC is the rank of jxi 0 j after discarding values of xi D 0 , and n is the number of xi values not
equal to 0 . Average ranks are used for tied values.
If n 20, the significance of S is computed from the exact distribution of S, where the distribution is a
convolution of scaled binomial distributions. When n > 20, the significance of S is computed by treating
r
n 1
S
nV S 2
as a Student t variate with n 1 degrees of freedom. V is computed as
1 1 X
V D n.n C 1/.2n C 1/ ti .ti C 1/.ti 1/
24 48
where the sum is over groups tied in absolute value and where ti is the number of values in the ith group
(Iman 1974, Conover 1980). The null hypothesis tested is that the mean (or median) is 0 , assuming that the
distribution is symmetric. Refer to Lehmann and D’Abrera (1975).
Tests for Normality

You can use the NORMALTEST option in the PROC CAPABILITY statement to request several tests of the
hypothesis that the analysis variable values are a random sample from a normal distribution. These tests,
which are summarized in the table labeled Tests for Normality, include the following:
Shapiro-Wilk test
Kolmogorov-Smirnov test
Anderson-Darling test
Cramér–von Mises test
Tests for normality are particularly important in process capability analysis because the commonly used
capability indices are difficult to interpret unless the data are at least approximately normally distributed.
Furthermore, the confidence limits for capability indices displayed in the table labeled Process Capability
Indices require the assumption of normality. Consequently, the tests of normality are always computed when
you specify the SPEC statement, and a note is added to the table when the hypothesis of normality is rejected.
You can specify the particular test and the significance level with the CHECKINDICES option.
Shapiro-Wilk Test
If the sample size is 2000 or less, the procedure computes the Shapiro-Wilk statistic W (also denoted as
Wn to emphasize its dependence on the sample size n). The statistic Wn is the ratio of the best estimator
of the variance (based on the square of a linear combination of the order statistics) to the usual corrected
sum of squares estimator of the variance. When n is greater than three, the coefficients to compute the linear
combination of the order statistics are approximated by the method of Royston (1992). The statistic Wn is
always greater than zero and less than or equal to one .0 < W 1/.
Small values of W lead to rejection of the null hypothesis. The method for computing the p-value (the
probability of obtaining a W statistic less than or equal to the observed value) depends on n. For n = 3,
the probability distribution of W is known and is used to determine the p-value. For n > 4, a normalizing
transformation is computed:

. log. log.1 Wn // /= if 4 n 11
Zn D
.log.1 Wn / /= if 12 n 2000
The values of , , and are functions of n obtained from simulation results. Large values of Zn indicate
departure from normality, and because the statistic Zn has an approximately standard normal distribution,
this distribution is used to determine the p-values for n > 4.
EDF Tests for Normality

The Kolmogorov-Smirnov, Anderson-Darling and Cramér–von Mises tests for normality are based on the
empirical distribution function (EDF) and are often referred to as EDF tests. EDF tests for a variety of
non-normal distributions are available in the HISTOGRAM statement; see the section “EDF Goodness-of-Fit
Tests” on page 364 for details. For a thorough discussion of these tests, refer to D’Agostino and Stephens
(1986).
The empirical distribution function is defined for a set of n independent observations X1 ; : : : ; Xn with a
common distribution function F .x/. Under the null hypothesis, F .x/ is the normal distribution. Denote the
observations ordered from smallest to largest as X.1/ ; : : : ; X.n/ . The empirical distribution function, Fn .x/,
is defined as
8
< 0; x < X.1/
i
Fn .x/ D ; X.i / x < X.i C1/ ; i D 1; : : : ; n 1
: n
1; X.n/ x
Note that Fn .x/ is a step function that takes a step of height n1 at each observation. This function estimates
the distribution function F .x/. At any value x, Fn .x/ is the proportion of observations less than or equal to x,
while F .x/ is the probability of an observation less than or equal to x. EDF statistics measure the discrepancy
between Fn .x/ and F .x/.
The EDF tests make use of the probability integral transformation U D F .X /. If F .X / is the distribution
function of X, the random variable U is uniformly distributed between 0 and 1. Given n observations
X.1/ ; : : : ; X.n/ , the values U.i / D F .X.i / / are computed. These values are used to compute the EDF test
statistics, as described in the next three sections. The CAPABILITY procedures computes the associated
p-values by interpolating internal tables of probability levels similar to those given by D’Agostino and
Stephens (1986).
Kolmogorov-Smirnov Test
The Kolmogorov-Smirnov statistic (D) is defined as
D D supx jFn .x/ F .x/j
The Kolmogorov-Smirnov statistic belongs to the supremum class of EDF statistics. This class of statistics is
based on the largest vertical difference between F .x/ and Fn .x/.
The Kolmogorov-Smirnov statistic is computed as the maximum of D C and D , where D C is the largest
vertical distance between the EDF and the distribution function when the EDF is greater than the distribution
function, and D is the largest vertical distance when the EDF is less than the distribution function.

D C D maxi ni U.i /

D D maxi U.i / i n1
D max D C ; D

D
PROC CAPABILITY uses a modified Kolmogorov D statistic to test the data against a normal distribution
with mean and variance equal to the sample mean and variance.
Anderson-Darling Test
The Anderson-Darling statistic and the Cramér–von Mises statistic belong to the quadratic class of EDF
statistics. This class of statistics is based on the squared difference .Fn .x/ F .x//2 . Quadratic statistics
have the following general form:
Z C1
QDn .Fn .x/ F .x//2 .x/dF .x/
1
The function .x/ weights the squared difference .Fn .x/ F .x//2 .
The Anderson-Darling statistic (A2 ) is defined as
Z C1
A2 D n .Fn .x/ F .x//2 ŒF .x/ .1 F .x// 1
dF .x/
1
1
Here the weight function is .x/ D ŒF .x/ .1 F .x// .
The Anderson-Darling statistic is computed as
n
1 X
A2 D

n .2i 1/ log U.i / C .2n C 1 2i / log f1 U.i /
n
i D1
Cramér–von Mises Test

The Cramér–von Mises statistic (W 2 ) is defined as
Z C1
2
W Dn .Fn .x/ F .x//2 dF .x/
1
Here the weight function is .x/ D 1.
The Cramér–von Mises statistic is computed as
n
2i 1 2

2
X 1
W D U.i / C
2n 12n
i D1
Percentile Computations
The CAPABILITY procedure automatically computes the 1st, 5th, 10th, 25th, 50th, 75th, 90th, 95th, and
99th percentiles (quantiles), as well as the minimum and maximum of each analysis variable. To compute
percentiles other than these default percentiles, use the PCTLPTS= and PCTLPRE= options in the OUTPUT
statement.
You can specify one of five definitions for computing the percentiles with the PCTLDEF= option. Let n be
the number of nonmissing values for a variable, and let x1 ; x2 ; : : : ; xn represent the ordered values of the
t
variable. Let the tth percentile be y, set p D 100 , and let
np D j C g when PCTLDEF=1, 2, 3, or 5
.n C 1/p D j C g when PCTLDEF=4
where j is the integer part of np, and g is the fractional part of np. Then the PCTLDEF= option defines the
tth percentile, y, as described in Table 6.7.
Table 6.7 PCTLDEF= Option Details
PCTLDEF= Description Formula

1 Weighted average at xnp y D .1 g/xj C gxj C1
where x0 is taken to be x1
y D xi if g ¤ 21
2 Observation numbered y D xj if g D 21 and j is even
closest to np y D xj C1 if g D 21 and j is odd
where i is the integer part of np C 12
y D xj if g D 0
3 Empirical distribution function
y D xj C1 if g > 0
4 Weighted average aimed y D .1 g/xj C gxj C1

at x.nC1/p where xnC1 is taken to be xn
y D 12 .xj C xj C1 / if g D 0
5 Empirical distribution function
y D xj C1 if g > 0
with averaging
Weighted Percentiles
When you use a WEIGHT statement, the percentiles are computed differently. The 100pth weighted percentile
y is computed from the empirical distribution function with averaging
8
ˆ
< x1 if w1 > pW
1
.xi C xi C1 / if ij D1 wj D pW
P
yD 2
if ij D1 wj < pW < ijC1
ˆ P P
: xi C1 D1 wj
where wi is the weight associated with xi , and where W D niD1 wi is the sum of the weights.
P
Note that the PCTLDEF= option is not applicable when a WEIGHT statement is used. However, in this
case, if all the weights are identical, the weighted percentiles are the same as the percentiles that would be
computed without a WEIGHT statement and with PCTLDEF=5.
Confidence Limits for Percentiles

You can use the CIPCTLNORMAL option to request confidence limits for percentiles which assume the
data are normally distributed. These limits are described in Section 4.4.1 of Hahn and Meeker (1991). When
0:0 < p < 0:5, the two-sided 100.1 ˛/% confidence limits for the 100p-th percentile are
lower limit D XN g 0 .˛=2I 1 p; n/s

0
upper limit D XN g .1 ˛=2I p; n/s
where n is the sample size. When 0:5 p < 1:0, the two-sided 100.1 ˛/% confidence limits for the
100p-th percentile are
lower limit D XN C g 0 .˛=2I 1 p; n/s

upper limit D XN C g 0 .1 ˛=2I p; n/s
One-sided 100.1 ˛/% confidence bounds are computed by replacing ˛=2 by ˛ in the appropriate preceding
equation. The factor g 0 . ; p; n/ is related to the noncentral t distribution and is described in Owen and Hua
(1977) and Odeh and Owen (1980).
You can use the CIPCTLDF option to request confidence limits for percentiles which are distribution free (in
particular, it is not necessary to assume that the data are normally distributed). These limits are described
in Section 5.2 of Hahn and Meeker (1991). The two-sided 100.1 ˛/% confidence limits for the 100p-th
percentile are
lower limit D X.l/

upper limit D X.u/
where X.j / is the jth order statistic when the data values are arranged in increasing order:
X.1/ X.2/ X.n/
The lower rank l and upper rank u are integers that are symmetric (or nearly symmetric) around bnpc C 1
where bnpc is the integer part of np, and where n is the sample size. Furthermore, l and u are chosen so that
X.l/ and X.u/ are as close to XbnpcC1 as possible while satisfying the coverage probability requirement
Q.u 1I n; p/ Q.l 1I n; p/ 1 ˛
where Q.kI n; p/ is the cumulative binomial probability

k
X n
Q.kI n; p/ D p i .1 p/n i
i
i D0
In some cases, the coverage requirement cannot be met, particularly when n is small and p is near 0 or 1. To
relax the requirement of symmetry, you can specify CIPCTLDF( TYPE = ASYMMETRIC ). This option
requests symmetric limits when the coverage requirement can be met, and asymmetric limits otherwise.
If you specify CIPCTLDF( TYPE = LOWER ), a one-sided 100.1 ˛/% lower confidence bound is computed
as Xl , where l is the largest integer that satisfies the inequality
1 Q.l 1I n; p/ 1 ˛
with 0 < l n. If you specify CIPCTLDF( TYPE = UPPER ), a one-sided 100.1 ˛/% upper confidence
bound is computed as Xu , where u is the smallest integer that satisfies the inequality
Q.u 1I n; p/ 1 ˛
where 0 < u n.
Note that confidence limits for percentiles are not computed when a WEIGHT statement is specified.
Robust Estimators
The CAPABILITY procedure provides several methods for computing robust estimates of location and scale,
which are insensitive to outliers in the data.
Winsorized Means
The k-times Winsorized mean is a robust estimator of location which is computed as
0 1
nX k 1
1
xN wk D @.k C 1/x.kC1/ C x.i / C .k C 1/x.n k/ A
n
i DkC2
where n is the number of observations, and x.i / is the ith order statistic when the observations are arranged in
increasing order:
x.1/ x.2/ x.n/
The Winsorized mean is the mean computed after replacing the k smallest observations with the (k + 1)st
smallest observation, and the k largest observations with the (k + 1)st largest observation.
For data from a symmetric distribution, the Winsorized mean is an unbiased estimate of the population mean.
However, the Winsorized mean does not have a normal distribution even if the data are normally distributed.
The Winsorized sum of squared deviations is defined as
nX
k 1
2 2
swk D .k C 1/.x.kC1/ xN wk / C .x.i / xN wk /2 C .k C 1/.x.n k/ xN wk /2
i DkC2
A Winsorized t test is given by

xN wk 0
twk D
STDERR.xN wk /
where the standard error of the Winsorized mean is
n 1 swk
STDERR.xN wk / D p
n 2k 1 n.n 1/
When the data are from a symmetric distribution, the distribution of twk is approximated by a Student’s t
distribution with n 2k 1 degrees of freedom. Refer to Tukey and McLaughlin (1963) and Dixon and
Tukey (1968).
A 100.1 ˛/% Winsorized confidence interval for the mean has upper and lower limits
xN wk ˙ t1 N wk /
˛=2 STDERR.x
where t1 ˛=2 is the .1 ˛=2/100th percentile of the Student’s t distribution with n 2k 1 degrees of
freedom.
Trimmed Means
The k-times trimmed mean is a robust estimator of location which is computed as
n
Xk
1
xN t k D x.i /
n 2k
i DkC1
where n is the number of observations, and x.i / is the ith order statistic when the observations are arranged in
increasing order:
x.1/ x.2/ x.n/
The trimmed mean is the mean computed after the k smallest observations and the k largest observations in
the sample are deleted.
For data from a symmetric distribution, the trimmed mean is an unbiased estimate of the population mean.
However, the trimmed mean does not have a normal distribution even if the data are normally distributed.
A robust estimate of the variance of the trimmed mean tt k can be obtained from the Winsorized sum of
squared deviations; refer to Tukey and McLaughlin (1963). the corresponding trimmed t test is given by
xN t k 0
tt k D
STDERR.xN t k /
where the standard error of the trimmed mean is
st k
STDERR.xN t k / D p
.n 2k/.n 2k 1/
and swk is the square root of the Winsorized sum of squared deviations.
When the data are from a symmetric distribution, the distribution of tt k is approximated by a Student’s t
distribution with n 2k 1 degrees of freedom. Refer to Tukey and McLaughlin (1963) and Dixon and
Tukey (1968).
A 100.1 ˛/% trimmed confidence interval for the mean has upper and lower limits
xN t k ˙ t1 Nt k /
˛=2 STDERR.x
where t1 ˛=2 is the .1 ˛=2/100th percentile of the Student’s t distribution with n 2k 1 degrees of
freedom.
Robust Estimates of Scale

The sample standard deviation, which is the most commonly used estimator of scale, is sensitive to outliers.
Robust scale estimators, on the other hand, remain bounded when a single data value is replaced by an
arbitrarily large or small value. The CAPABILITY procedure computes several robust measures of scale,
including the interquartile range Gini’s mean difference G, the median absolute deviation about the median
(MAD), Qn , and Sn . In addition, the procedure computes estimates of the normal standard deviation
derived from each of these measures.
The interquartile range (IQR) is simply the difference between the upper and lower quartiles. For a normal
population, can be estimated as IQR/1.34898.
Gini’s mean difference is computed as

1 X
GD jxi xj j
n i <j
2
p p
For a normal population, the expected value of G is 2= . Thus G =2 is a robust estimator of when
the data are from a normal sample. For the normal distribution, this estimator has high efficiency relative to
the usual sample standard deviation, and it is also less sensitive to the presence of outliers.
A very robust scale estimator is the MAD, the median absolute deviation from the median (Hampel 1974),
which is computed as
MAD D medi .jxi medj .xj /j/
where the inner median, medj .xj /, is the median of the n observations, and the outer median (taken over i)
is the median of the n absolute values of the deviations about the inner median. For a normal population,
1.4826MAD is an estimator of .
The MAD has low efficiency for normal distributions, and it may not always be appropriate for symmetric
distributions. Rousseeuw and Croux (1993) proposed two statistics as alternatives to the MAD. The first is
Sn D 1:1926 medi .medj .jxi xj j//
where the outer median (taken over i) is the median of the n medians of jxi xj j, j D 1; 2; : : : ; n. To
reduce small-sample bias, csn Sn is used to estimate , where csn is a correction factor; refer to Croux and
Rousseeuw (1992).
The second statistic is
Qn D 2:2219 fjxi xj jI i < j g.k/
where

h
kD
2

n
and h D Œn=2 C 1. In other words, Qn is 2.2219 times the kth order statistic of the distances between
2
the data points. The bias-corrected statistic cq n Qn is used to estimate , where cq n is a correction factor;
refer to Croux and Rousseeuw (1992).
Computing the Mode

The mode is the value that occurs most often in a set of observations. The CAPABILITY procedure counts
repetitions of the actual values (or the rounded values, if you specify the ROUND= option). If a tie occurs for
the most frequent value, the procedure reports the lowest mode in the table labeled Basic Statistical Measures.
To list all possible modes, specify the MODES option in the PROC CAPABILITY statement. When no
repetitions occur in the data, the procedure does not report the mode. The WEIGHT statement has no effect
on the mode.
Assumptions and Terminology for Capability Indices

One of the fundamental assumptions in process capability analysis is that the process must be in statistical
control. Without statistical control, the process is not predictable, the concept of a process distribution does
not apply, and quantities related to the distribution, such as probabilities, percentiles, and capability indices,
cannot be meaningfully estimated. Additionally, all of the standard process capability indices described in
the next section require that the process distribution be normal, or at least approximately normal.
In many industries, statistical control is routinely checked with a Shewhart chart (such as an XN and R chart)
before capability indices such as

USL LSL
Cpk D min ;
3 3
are computed. The control chart analysis yields estimates for the process mean and standard deviation ,
which are based on subgrouped data and can be used to estimate Cpk . In particular, can be estimated by
N 2
sR D R=d
rather than the ungrouped sample standard deviation

sX
1
sD n.xi x/N 2
n 1
i D1
You can use the SHEWHART procedure to carry out the control chart analysis and to compute capability
indices based on sR . On the other hand, the CAPABILITY procedure computes indices based on s.
Some industry manuals distinguish these two approaches. For instance, the ASQC/AIAG manual Fundamen-
tal Process Control uses the notation Cpk for the estimate based on sR , and it uses the notation Ppk for the
estimate based on s. However, assuming that the process is in control and only common cause variation is
present, both sR and s are estimates of the same parameter , and so there is fundamentally no difference in
the two approaches2 .
Once control has been established, attention should focus on the distribution of the process measurements,
and at this point there is no practical or statistical advantage to working with subgrouped measurements.
In fact, the use of s is closely associated with a wide variety of methods that are highly useful for process
capability analysis, including tests for normality, graphical displays such as histograms and probability plots,
and confidence intervals for parameters and capability indices.
Standard Capability Indices

This section provides computational details for the standard process capability indices computed by the
CAPABILITY procedure: Cp , CPL, CPU , Cpk , and Cpm .
The Index Cp
The process capability index Cp , sometimes called the “process potential index,” the “process capability
ratio,” or the “inherent capability index,” is estimated as
b p D USL
C
LSL
6s
2 Statistically, s is a more efficient estimator of than sR .
where USL is the upper specification limit, LSL is the lower specification limit, and s is the sample standard
deviation. If you do not specify both the upper and the lower specification limits in the SPEC statement or
the SPEC= data set, then Cp is assigned a missing value.
The interpretation of Cp can depend on the application, on past experience, and on local practice. However,
broad guidelines for interpretation have been proposed by several authors. Ekvall and Juran (1974) classify
Cp values as
“not adequate” if Cp < 1
“adequate” if 1 Cp 1:33, but requiring close control as Cp approaches 1
“more than adequate” if Cp > 1:33
Montgomery (1996) recommends minimum values of Cp as
1.33 for existing processes
1.50 for new processes or for existing processes when the variable is critical (for example, related to
safety or strength)
1.67 for new processes when the variable is critical
Exact 100.1 ˛/% lower and upper confidence limits for Cp (denoted by LCL and UCL) are computed
using percentiles of the chi-square distribution, as indicated by the following equations:
q
lower limit D COp 2˛=2;n 1
=.n 1/
q
upper limit D COp 21 ˛=2;n 1
=.n 1/
Here, 2˛; denotes the lower 100˛th percentile of the chi-square distribution with degrees of freedom.
Refer to Chou, Owen, and Borrego (1990) and Kushler and Hurley (1992).
You can specify ˛ with the ALPHA= option in the PROC CAPABILITY statement or with the CIINDICES(
ALPHA=value ) in the PROC CAPABILITY statement. The default value is 0.05. You can save these limits
in the OUT= data set by specifying the keywords CPLCL and CPUCL in the OUTPUT statement. In addition,
you can display these limits on plots produced by the CAPABILITY procedure by specifying the keywords
in the INSET statement.
The Index CPL

The process capability index CPL is estimated as
b
CPL D
XN LSL
3s
where XN is the sample mean, LSL is the lower specification limit, and s is the sample standard deviation. If
you do not specify the lower specification limit in the SPEC statement or the SPEC= data set, then CPL is
assigned a missing value.
Montgomery (1996) refers to CPL as the “process capability ratio” in the case of one-sided lower specifica-
tions and recommends minimum values as follows:
1.25 for existing processes

1.45 for new processes or for existing processes when the variable is critical
1.60 for new processes when the variable is critical
Exact 100.1 ˛/% lower and upper confidence limits for CPL are computed using a generalization of the
method of Chou, Owen, and Borrego (1990), who point out that the 100.1 ˛/ lower confidence limit for
CPL (denoted by CPLLCL )satisfies the equation
p p
PrfTn 1 .ı D 3 n/ CPLLCL 3CPL ng D 1 ˛
where Tn 1 .ı/ has a non-central t distribution with n – 1 degrees of freedom and noncentrality parameter ı.
You can specify ˛ with the ALPHA= option in the PROC CAPABILITY statement. The default value is 0.05.
The confidence limits can be saved in an output data set by specifying the keywords CPLLCL and CPLUCL
in the OUTPUT statement. In addition, you can display these limits on plots produced by the CAPABILITY
procedure by specifying these keywords in the INSET statement.
The Index CPU

The process capability index CPU is estimated as
1
CPU D
USL XN
3s
where USL is the upper specification limit, XN is the sample mean, and s is the sample standard deviation. If
you do not specify the upper specification limit in the SPEC statement or the SPEC= data set, then CPU is
Montgomery (1996) refers to CPU as the “process capability ratio” in the case of one-sided upper specifica-
tions and recommends minimum values that are the same as those specified previously for CPL.
Exact 100.1 ˛/% lower and upper confidence limits for CPU are computed using a generalization of the
method of Chou, Owen, and Borrego (1990), who point out that the 100.1 ˛/ lower confidence limit for
CPU (denoted by CPULCL )satisfies the equation
p p
PrfTn 1 .ı D 3 n CPULCL 3CPU ng D 1 ˛
where Tn 1 .ı/ has a non-central t distribution with n – 1 degrees of freedom and noncentrality parameter ı.
You can specify ˛ with the ALPHA= option in the PROC CAPABILITY statement. The default value is 0.05.
The confidence limits can be saved in an output data set by specifying the keywords CPULCL and CPUUCL
in the OUTPUT statement. In addition, you can display these limits on plots produced by the CAPABILITY
procedure by specifying these keywords in the INSET statement.
The Index Cpk

The process capability index Cpk is defined as
1
Cpk D min.USL ; LSL/ D min.CPU ; CPL/
3
Note that the indices Cpk , Cp , and k are related as Cpk D Cp .1 k/. The CAPABILITY procedure estimates
Cpk as
b pk D 1 min.USL
C XN ; XN LSL/ D min.CPU ; CPL/
3s
where USL is the upper specification limit, LSL is the lower specification limit, XN is the sample mean, and s
is the sample standard deviation.
If you specify only the upper limit in the SPEC statement or the SPEC= data set, then Cpk is computed as
CPU , and if you specify only the lower limit in the SPEC statement or the SPEC= data set, then Cpk is
computed as CPL.
Bissell (1990) derived approximate two-sided 95% confidence limits for Cpk by assuming that the distribution
b pk is normal. Using Bissell’s approach, 100.1 ˛/% lower and upper confidence limits can be computed
of C
as
" s #
1 1 1
lower limit D C
b pk 1 ˆ .1 ˛=2/ C
9nCb2 2.n 1/
pk
" s #
1 1 1
upper limit D C
b pk 1 C ˆ .1 ˛=2/ C
b2
9nC 2.n 1/
pk
where ˆ denotes the cumulative standard normal distribution function. Kushler and Hurley (1992) concluded
that Bissell’s method gives reasonably accurate results.
You can specify ˛ with the ALPHA= option in the PROC CAPABILITY statement. The default value is
0.05. These limits can be saved in an output data set by specifying the keywords CPKLCL and CPKUCL in
the OUTPUT statement. In addition, you can display these limits on plots produced by the CAPABILITY
procedure by specifying these same keywords in the INSET statement.
The Index Cpm

The process capability index Cpm is intended to account for deviation from the target T in addition to
variability from the mean. This index is often defined as
USL LSL
Cpm D p
6 2 C . T /2
A closely related version of Cpm is the index
min .USL T; T LSL/ d jT mj

Cpm D p D p
3 2 C . T /2 3 2 C . T /2
where d D .USL LSL/=2 and m D .USL C LSL/=2. If T D m, then Cpm D Cpm . However, if T ¤ m,
then both indices suffer from problems of interpretation, as pointed out by Kotz and Johnson (1993), and
their use should be avoided in this case.
The CAPABILITY procedure computes an estimator of Cpm as
b pm D min.USL T; T LSL/
C p
3 s 2 C .XN T /2
where s is the sample standard deviation.
If you specify only a single specification limit SL in the SPEC statement or the SPEC= data set, then Cpm is
estimated as
b pm D p jT SLj
C
3 s 2 C .XN T /2
Boyles (1991) proposed a slightly modified point estimate for Cpm computed as
e pm D q .USL LSL/=2
C
3 . n n 1 /s 2 C .XN T /2
Boyles also suggested approximate two-sided 100.1 ˛/% confidence limits for Cpm , which are computed
as
q
lower limit D C e pm 2 =
˛=2;
q
upper limit D C e pm 2 =
1 ˛=2;
Here 2˛; denotes the lower 100˛th percentile of the chi-square distribution with degrees of freedom,
where equals
N
n.1 C . X s T /2 /
N
1 C 2. X s T /2
You can specify ˛ with the ALPHA= option in the PROC CAPABILITY statement. The default value is
0.05. These confidence limits can be saved in an output data set by specifying the keywords CPMLCL and
CPMUCL in the OUTPUT statement. In addition, you can display these limits on plots produced by the
CAPABILITY procedure by specifying these keywords in the INSET statement.
Specialized Capability Indices

This section describes a number of specialized capability indices which you can request with the SPE-
CIALINDICES option in the PROC CAPABILITY statement.
The Index k
The process capability index k (also denoted by K) is computed as
2jm XN j
kD
USL LSL
where m D 12 .USL C LSL/ is the midpoint of the specification limits, XN is the sample mean, USL is the
upper specification limit, and LSL is the lower specification limit.
The formula for k used here is given by Kane (1986). Note that k is sometimes computed without taking the
absolute value of m XN in the numerator. See Wadsworth, Stephens, and Godfrey (1986).
If you do not specify the upper and lower limits in the SPEC statement or the SPEC= data set, then k is
C
Boyles’ Index Cpm
C which is defined as
Boyles (1992) proposed the process capability index Cpm
" # 1=2
1 EX <T .X T /2 EX>T .X T /2

C
Cpm D C
3 .T LSL/2 .USL T /2
He proposed this index as a modification of Cpm for use when ¤ T . The quantities
EX <T .X T /2 D E .X T /2 jX < T P r ŒX < T

and
T /2 D E .X T /2 jX > T P r ŒX > T

EX >T .X
are referred to as semivariances. Kotz and Johnson (1993) point out that if T D .LSL C USL/=2, then
C DC
Cpm pm .
C is
Kotz and Johnson (1993) suggest that a natural estimator for Cpm
2 (P 3
2
P 2
) 1=2
1 41 Xi <T .X i T / Xi >T .Xi T /
bC
C pm D C 5
3 n .T LSL/2 .USL T /2
Note that this index is not defined when either of the specification limits is equal to the target T. Refer to
Section 3.5 of Kotz and Johnson (1993) for further details.
The Index Cjkp

Johnson, Kotz, and Pearn (1994) introduced a so-called “flexible” process capability index which takes into
account possible differences in variability above and below the target T. They defined this index as
!
1 USL T T LSL
Cjkp D p min p ;p
3 2 EX >T Œ.X T /2 EX <T Œ.X T /2
where d D .USL LSL/=2.

A natural estimator of this index is
0 1
1
b jkp D p USL T T LSL
C min @ qP ; qP
B C
A
3 2 .Xi T /2 =n .Xi T /2 =n
Xi >T Xi <T
For further details, refer to Section 4.4 of Kotz and Johnson (1993).
The Indices Cpm .a/

The class of capability indices Cpm .a/, indexed by the parameter a (a > 0) allows flexibility in choosing
between the relative importance of variability and deviation of the mean from the target value T.
The class defined as
Cpm .a/ D .1 a 2 /Cp

where D . T /= . The motivation for this definition is that if jj is small, then
1 2
Cpm .1 /Cp
2
A natural estimator of Cpm .a/ is
8 !2 9
d b < N
X T =
C pm .a/ D 1 a
3s : s ;
where d D .USL LSL/=2. You can specify the value of a with the SPECIALINDICES(CPMA=) option in
the PROC CAPABILITY statement. By default, a = 0.5.
This index is not recommended for situation in which the target T is not equal to the midpoint of the
specification limits.
For additional details, refer to Section 3.7 of Kotz and Johnson (1993).
The Index Cp.5:15/

Johnson, Kotz, and Kemp (1992) suggest the class of process capability indices defined as
USL LSL
Cp. / D

where is chosen so that the proportion of conforming items is robust with respect to the shape of the process
distribution. In particular, Kotz and Johnson (1993) recommend use of
USL LSL
Cp.5:15/ D
5:15
which is estimated as
b p.5:15/ D USL LSL

C
5:15s
For details, refer to Section 4.3.2 of Kotz and Johnson (1993).
The Index Cpk .5:15/

Similarly, Kotz and Johnson (1993) recommend use of the robust capability index
d j .USL C LSL/=2j
Cpk .5:15/ D
2:575
where d D .USL LSL/=2. This index is estimated as
b pk .5:15/ D d jXN .USL C LSL/=2j

C
2:575s
For details, refer to Section 4.3.2 of Kotz and Johnson (1993).

The Index Cpmk

Pearn, Kotz, and Johnson (1992) proposed the index Cpmk
.USL LSL/=2 j mj
Cpmk D p
3 2 C . T /2
where m D .LCL C UCL/=2. A natural estimator for Cpmk is
b pmk D .USL LSL/=2 jXN mj

C q
3 . n n 1 /s 2 C .XN T /2
where m D .USL C LSL/=2.

For further details, refer to Section 3.6 of Kotz and Johnson (1993).
Wright’s Index Cs
Wright (1995) defines the capability index
min .USL ; LSL/
Cs D p
3 2 C . T /2 C 3 =
where 3 D E.X /3 .

A natural estimator of Cs is
N
b s D q .USL LSL/=2 jX mj
C
3 n n 1 s 2 C .XN T /2 C jc4 s 2 b3 j

where c4 is an unbiasing constant for the sample standard deviation, and b3 is a measure of skewness. Wright
(1995) shows that Cs compares favorably with Cpmk even when skewness is not present, and he advocates
the use of Cs for monitoring near-normal processes when loss of capability typically leads to asymmetry.
Chen and Kotz (1996) proposed a modification to Wright’s Cs index which introduces a multiplier, > 0,
and is estimated as
N
b s D q .USL LSL/=2 jX mj
C
3 n n 1 s 2 C .XN T /2 C jc4 s 2 b3 j

If you specify a value for with the SPECIALINDICES(CSGAMMA=) option, the index Cs is computed
with this modification. Otherwise it is computed using Wright’s original definition.
The Index Sjkp

Boyles (1994) proposed a smooth version of Cjkp defined as
!
USL T T LSL
Sjkp D S p ;p
2EX >T Œ.X T /2 2EX <T Œ.X T /2
The CAPABILITY procedure estimates Sjkp as
0 1
USL T T LSL
S jkp D S @ q P ;q P
b B C
A
2 Xi >T .Xi T /2 =n 2 Xi <T .Xi T /2 =n
where S.x; y/ D ˆ 1 Œfˆ.x/ C ˆ.y/g=2=3.
The Index Cpp

index. The first term measures inaccuracy
Chen (1998) devised a process incapability index based on the Cpm
and the second measures imprecision. The Cpp index is estimated as
!2 2
XN T

b pp D s
C C
d =3
d =3
where d D min.USL T; T LSL/.

00
The Index Cpp
The index Cpp does not handle asymmetric tolerances well, as discussed by Kotz and Lovelace (1998). To
00
address that shortcoming, Chen (1998) defined the index Cpp , which is estimated by
!2
00
b pp D A
b s
C
C
d =3 d =3
where
( )
.XN T /d .T XN /d
A D max
b ;
T LSL USL T
and d D .USL LSL/=2.
The Index Cpg

Marcucci and Beazley (1988) defined the index
1
Cpg D 2
Cpm
which is estimated as
b pg D 1
C
b 2pm
C
The Index Cpq

Gupta and Kotz (1997) introduced the index Cpq , which is estimated by
2 !2 3
b pq D C 1 XN T
C b p 41 5
2 s
The Index CpW

Bai and Choi (1997) defined the index
Cp
CpW D p
1 C j1 2Px j
where Px D Pr.X /. It is estimated by
bW C
bp
C p D q
1 C j1 bx j
2P
where P N For more information about CpW , see Kotz

b x is the fraction of observations less than or equal to X.
and Lovelace (1998).
W
The Index Cpk
Bai and Choi (1997) also proposed the index
( )
W USL LSL
Cpk D min p ; p
3 2Px 3 2.1 Px /
It is estimated by
8 9
< USL XN
ˆ
XN LSL =
>
bW
C pk D min q ; q
ˆ b x 3s 2.1 P
: 3s 2P bx / >
;
where P N For more information about C W , see Kotz

b x is the fraction of observations less than or equal to X.
pk
W
The Index Cpm
W , also introduced by Bai and Choi (1997), is defined as
The index Cpm
W Cpm
Cpm Dp
1 C j1 2PT j
where PT D Pr.X T /. It is estimated by
bW C
b pm
C pm D q
1 C j1 bT j
2P
where P W , see Kotz

b T is the fraction of observations less than or equal to T. For more information about Cpm
The Index Cpc

Luceño (1996) proposed the index
USL LSL
Cpc D q
6 2 EjX M j
where M D .USL C LSL/=2. It is estimated by
b pc D USLq LSL
C
6 2 c
where
n
1X
cD jXi Mj
n
i D1
Vännmann’s Index Cp .u; v/

Vännmann (1995) introduced the generalized index Cp .u; v/, which reduces to the following capability
indices given appropriate choices of u and v:
Cp .0; 0/ D Cp
Cp .0; 1/ D Cpk
Cp .1; 0/ D Cpm
Cp .1; 1/ D Cpmk
Cp .u; v/ is defined as
d uj M j
Cp .u; v/ D p
3 2 C v. T /2
and estimated by
N
b p .u; v/ D q d ujX M j
C
3 . n n 1 /s 2 C v.XN T /2
You can specify u with the SPECIALINDICES(CPU=) option and v with the SPECIALINDICES(CPV=)
option. By default, u = 0 and v = 4.
Vännmann’s Index Cp .v/

Vännmann (1997) also proposed the index Cp .v/, which is equivalent to Cp .u; v/ with u = 1. It is estimated
as
N
b p .v/ D q d jX M j
C
3 . n n 1 /s 2 C v.XN T /2
You can specify v with the SPECIALINDICES(CPV=) option. By default, v = 4.
Missing Values
If a variable for which statistics are calculated has a missing value, that value is ignored in the calculation of
statistics, and the missing values are tabulated separately. A missing value for one such variable does not
affect the treatment of other variables in the same observation.
If the WEIGHT variable has a missing value, the observation is excluded from the analysis. If the FREQ
variable has a missing value, the observation is excluded from the analysis. If a variable in a BY or ID
statement has a missing value, the procedure treats it as it would treat any other value of a BY or ID variable.
ODS Tables
This section describes the ODS tables produced by the CAPABILITY procedure.
Table 6.8 summarizes the ODS tables that you can request with options in the PROC CAPABILITY statement.
Table 6.8 ODS Tables Produced with the PROC CAPABILITY

Statement
Table Name Description Option

BasicIntervals Confidence intervals for mean, CIBASIC
standard deviation, variance
BasicMeasures Measures of location and variability Default
ExtremeObs Extreme observations Default
ExtremeValues Extreme values NEXTRVAL=
Frequencies Frequencies FREQ
LocationCounts Counts used for sign test and signed LOCCOUNTS
rank test
MissingValues Missing values Default
Modes Modes MODES
Moments Sample moments Default
Quantiles Quantiles Default
RobustScale Robust measures of scale ROBUSTSCALE
TestsForLocation Tests for location Default
TestsForNormality Tests for normality NORMALTEST
TrimmedMeans Trimmed means TRIMMED=
WinsorizedMeans Winsorized means WINSORIZED=
Table 6.9 summarizes the ODS tables related to capability indices that you can request with options in the
PROC CAPABILITY statement when you provide specification limits with a SPEC statement or with a
SPEC= data set.
Table 6.9 ODS Tables Related to Specification Limits

CIProbExSpecs Confidence limits for probabilities of CIPROBEX
exceeding specifications
Indices Standard capability indices Default
SpecialIndices Specialized capability indices SPECIALINDICES
Specifications Percents outside specification limits Default
based on empirical
Table 6.10 summarizes the ODS tables related to fitted distributions that you can request with options in the
HISTOGRAM statement.
Table 6.10 ODS Tables Produced with the HISTOGRAM

Statement

Bins Histogram bins MIDPERCENTS suboption with any
distribution option, such as
NORMAL(MIDPERCENTS)
FitIndices Capability indices INDICES suboption with any
computed from fitted distribution option, such as
distribution LOGNORMAL(INDICES)
FitQuantiles Quantiles of fitted Any distribution option such as
distribution NORMAL
GoodnessOfFit Goodness-of-fit tests for Any distribution option such as
fitted distribution NORMAL
ParameterEstimates Parameter estimates for Any distribution option such as
Specifications Percents outside Any distribution option such as
specification limits based NORMAL
on empirical and fitted
distributions
Table 6.11 summarizes the ODS tables that you can request with options in the INTERVALS statement.
Table 6.11 ODS Tables Produced with the INTERVALS Statement

Intervals1 Prediction interval for future observations METHODS=1
Intervals2 Prediction interval for mean METHODS=2
Intervals3 Tolerance interval for proportion of population METHODS=3
Intervals4 Confidence limits for mean METHODS=4
Intervals5 Prediction interval for standard deviation METHODS=5
Intervals6 Confidence limits for standard deviation METHODS=6
Examples: CAPABILITY Procedure

This section provides a more advanced example of the PROC CAPABILITY statement.
Example 6.1: Reading Specification Limits

N OTE : See Reading Spec Limits from an Input Data Set in the SAS/QC Sample Library.
You can specify specification limits either in the SPEC statement or in a SPEC= data set. In “Computing
Capability Indices” on page 209, limits were specified in a SPEC statement. This example illustrates how
to create a SPEC= data set to read specification limits with the SPEC= option in the PROC CAPABILITY
statement.
Consider the drink can data presented in “Computing Descriptive Statistics” on page 207. Suppose, in
addition to the fluid weight of each drink can, the weight of the can itself is stored in a variable named
Cweight, and both variables are saved in a data set called Can2. A partial listing of Can2 follows:
proc print data=Can2(obs=5);

run;
Output 6.1.1 The Data Set Can2
Obs Weight Cweight

1 12.07 1.07
2 12.02 0.86
3 12.00 1.06
4 12.01 1.08
5 11.98 1.02
The following DATA step creates a data set named Limits containing specification limits for the fluid weight
and the can weight. Limits has 4 variables (_VAR_, _LSL_, _USL_, and _TARGET_) and 2 observations. The
first observation contains the specification limit information for the variable Weight, and the second contains
the specification limit information for the variable Cweight.
data Limits;
length _var_ $8;
_var_ = 'Weight';
_lsl_ = 11.95;
_target_ = 12;
_usl_ = 12.05;
output;
_var_ = 'Cweight';
_lsl_ = 0.90;
_target_ = 1;
_usl_ = 1.10;
output;
run;
The following statements read the specification information from the Limits data set into the CAPABILITY
procedure by using the SPEC= option. These statements print summary statistics, capability indices, and
specification limit information for Weight and Cweight. Figure 6.1 and Figure 6.2 display the output for
Weight. Output 6.1.2 displays the output for Cweight.
Examples: CAPABILITY Procedure F 261
title 'Process Capability Analysis of Drink Can Data';

proc capability data=Can2 specs=Limits;
var Cweight;
run;
Output 6.1.2 Printed Output for Variable Cweight

Process Capability Analysis of Drink Can Data

Variable: Cweight (Can Weight (ounces))
Moments
N 100 Sum Weights 100
Mean 1.004 Sum Observations 100.4
Std Deviation 0.06330941 Variance 0.00400808
Skewness -0.074821 Kurtosis -0.5433858
Uncorrected SS 101.1984 Corrected SS 0.3968
Coeff Variation 6.30571767 Std Error Mean 0.00633094
Basic Statistical Measures

Location Variability
Mean 1.004000 Std Deviation 0.06331
Median 1.000000 Variance 0.00401
Mode 1.040000 Range 0.29000
Interquartile Range 0.08500
Note: The mode displayed is the smallest of 2 modes with a count of 8.
Tests for Location: Mu0=0

Student's t t 158.5862 Pr > |t| <.0001
Sign M 50 Pr >= |M| <.0001
Signed Rank S 2525 Pr >= |S| <.0001
Tests for Normality

Shapiro-Wilk W 0.987310 Pr < W 0.4588
Kolmogorov-Smirnov D 0.061410 Pr > D >0.1500
Cramer-von Mises W-Sq 0.048175 Pr > W-Sq >0.2500
Output 6.1.2 continued
Quantiles (Definition 5)
Level Quantile
100% Max 1.150
99% 1.140
95% 1.105
90% 1.080
75% Q3 1.045
50% Median 1.000
25% Q1 0.960
10% 0.910
5% 0.900
1% 0.870
0% Min 0.860
Extreme Observations
Lowest Highest
Value Obs Value Obs
0.86 2 1.11 42
0.88 89 1.12 28
0.88 64 1.12 34
0.90 68 1.13 48
0.90 59 1.15 52
Specification Limits
Limit Percent
Lower (LSL) 0.900000 % < LSL 3.00000
Target 1.000000 % Between 92.00000
Upper (USL) 1.100000 % > USL 5.00000

Cp 0.526515 0.453237 0.599670
CPL 0.547575 0.446607 0.647299
CPU 0.505454 0.408856 0.600808
Cpk 0.505454 0.409407 0.601501
Cpm 0.525467 0.454973 0.601113
Example 6.2: Enhancing Reference Lines

N OTE : See Controlling the Appearance of Spec Limits in the SAS/QC Sample Library.
A telecommunications company manufactures amplifiers to be used in telephones. Each amplifier is designed
to boost the input signal by 5 decibels (dB). Because it is difficult to make every amplifier’s boosting power
exactly 5 decibels, the company decides that amplifiers that boost the input signal between 4 and 6 decibels
are acceptable. Therefore, the target value is 5 decibels, and the lower and upper specification limits are 4 and
6 decibels, respectively. The following data set contains the boosting powers of a sample of 75 amplifiers:
data Amps;
label Decibels = 'Amplification in Decibels (dB)';
input Decibels @@;
datalines;
4.54 4.87 4.66 4.90 4.68 5.22 4.43 5.14 3.07 4.22
5.09 3.41 5.75 5.16 3.96 5.37 5.70 4.11 4.83 4.51
4.57 4.16 5.73 3.64 5.48 4.95 4.57 4.46 4.75 5.38
5.19 4.35 4.98 4.87 3.53 4.46 4.57 4.69 5.27 4.67
5.03 4.50 5.35 4.55 4.05 6.63 5.32 5.24 5.73 5.08
5.07 5.42 5.05 5.70 4.79 4.34 5.06 4.64 4.82 3.24
4.79 4.46 3.84 5.05 5.46 4.64 6.13 4.31 4.81 4.98
4.95 5.57 4.11 4.15 5.95
;
The SPEC statement provides several options to control the appearance of reference lines for the specification
limits and the target value. The following statements use the data set Amps to create a histogram that
demonstrates some of these options:
ods graphics off;

legend2 FRAME CFRAME=ligr CBORDER=black POSITION=center;
title 'Boosting Power of Telephone Amplifiers';
proc capability data=Amps;
spec target = 5 lsl = 4 usl = 6
ltarget = 2 llsl = 3 lusl = 4
wtarget = 2 wlsl = 2 wusl = 2
cleft cright;
histogram Decibels / cbarline = black;
run;
The resulting histogram is shown in Output 6.2.1. The LTARGET=, LLSL=, and LUSL= options control
the line type of the reference lines for the target, lower specification limit, and upper specification limit,
respectively. Likewise, the WTARGET=, WLSL=, and WUSL= options control the line widths. The
CLEFT= option controls the color used to fill the area to the left of the lower specification limit. Similarly,
the CRIGHT= option controls the color used to fill the area to the right of the upper specification limit.
Output 6.2.1 Controlling the Appearance of Specification Limits
Example 6.3: Displaying a Confidence Interval for Cpk

N OTE : See Displaying a Confidence Interval for Cpm in the SAS/QC Sample Library.
In this example, the capability index Cpk is computed for the amplification data in Amps. To examine the
accuracy of this estimate, the following statements calculate a 90% confidence interval for Cpk , then display
the interval on a histogram (shown in Output 6.3.1) with the INSET statement:
title 'Boosting Power of Telephone Amplifiers';

proc capability data=Amps noprint alpha=0.10;
var Decibels;
ltarget = 2 llsl = 3 lusl = 4;
histogram Decibels / odstitle = title;;
inset cpklcl cpk cpkucl / header = '90% Confidence Interval'
format = 6.3;
run;
The ALPHA= option in the PROC CAPABILITY statement controls the level of the confidence interval.
In this case, the 90% confidence interval on Cpk is wide (from 0.328 to 0.496), indicating that the process
may need adjustments in order to improve process variability. Confidence limits for capability indices can
be displayed using the INSET statement (as shown in Output 6.3.1) or saved in an output data set by using
the OUTPUT statement. For formulas and details about capability indices, see the section “Specialized
Capability Indices” on page 250. For more information about the INSET statement, see “INSET Statement:
CAPABILITY Procedure” on page 398.
Output 6.3.1 Confidence Interval on Cpk
The following statements can be used to produce a table of process capability indices including the index
Cpk :
ods select indices;

proc capability data=Amps alpha=0.10;
ltarget = 2 llsl = 3 lusl = 4;
var Decibels;
run;
Output 6.3.2 Process Capability Indices

Boosting Power of Telephone Amplifiers

Variable: Decibels (Amplification in Decibels (dB))

Cp 0.508962 0.439538 0.576922
CPL 0.411920 0.326620 0.495136
CPU 0.606004 0.501261 0.708127
Cpk 0.411920 0.327599 0.496241
Cpm 0.488674 0.425292 0.556732
CDFPLOT Statement: CAPABILITY Procedure
Overview: CDFPLOT Statement

The CDFPLOT statement plots the observed cumulative distribution function (cdf ) of a variable, defined as
FN .x/ D percent of nonmissing values x

number of values x
D 100%
N
where N is the number of nonmissing observations. The cdf is an increasing step function that has a vertical
jump of N1 at each value of x equal to an observed value. The cdf is also referred to as the empirical
cumulative distribution function (ecdf).
You can use options in the CDFPLOT statement to do the following:
superimpose specification limits
superimpose fitted theoretical distributions
specify graphical enhancements (such as color or text height)
You can also create a comparative cdf plot by using the CDFPLOT statement in conjunction with a CLASS
statement.
You have three alternatives for producing cdf plots with the CDFPLOT statement:

Getting Started: CDFPLOT Statement F 267
statement.
Getting Started: CDFPLOT Statement

Creating a Cumulative Distribution Plot
N OTE : See CDF Plot with Superimposed Normal Curve in the SAS/QC Sample Library.
This section introduces the CDFPLOT statement with a simple example. A company that produces fiber optic
cord is interested in the breaking strength of the cord. The following statements create a data set named Cord,
which contains 50 breaking strengths measured in pounds per square inch (psi), and they display the cdf plot
in Figure 6.5. The plot shows a symmetric distribution with observations concentrated 6.9 and 7.1. The plot
also shows that only a small percentage (< 5%) of the observations are below the lower specification limit of
6.8.
data Cord;
label Strength="Breaking Strength (psi)";
input Strength @@;
datalines;
6.94 6.97 7.11 6.95 7.12 6.70 7.13 7.34 6.90 6.83
7.06 6.89 7.28 6.93 7.05 7.00 7.04 7.21 7.08 7.01
7.05 7.11 7.03 6.98 7.04 7.08 6.87 6.81 7.11 6.74
6.95 7.05 6.98 6.94 7.06 7.12 7.19 7.12 7.01 6.84
6.91 6.89 7.23 6.98 6.93 6.83 6.99 7.00 6.97 7.01
;
title 'Cumulative Distribution Function of Breaking Strength';

proc capability data=Cord noprint;
spec lsl=6.8;
cdf Strength / odstitle=title;
run;
Figure 6.5 Cumulative Distribution Function
Syntax: CDFPLOT Statement

The syntax for the CDFPLOT statement is as follows:
CDFPLOT < variables > < / options > ;
You can specify the keyword CDF as an alias for CDFPLOT. You can specify any number of CDFPLOT
statements after a PROC CAPABILITY statement. The components of the CDFPLOT statement are described
as follows:
variables
specify variables for which to create cdf plots. If you specify a VAR statement, the variables must also
be listed in the VAR statement. Otherwise, the variables can be any numeric variables in the input
data set. If you do not specify variables in a CDFPLOT statement, then a cdf plot is created for each
variable listed in the VAR statement, or for each numeric variable in the input data set if you do not use
a VAR statement.
For example, suppose a data set named steel contains exactly three numeric variables, length, width
and height. The following statements create a cdf plot for each of the three variables:
Syntax: CDFPLOT Statement F 269
proc capability data=steel;

cdfplot;
run;
The following statements create a cdf plot for length and a cdf plot for width:

var length width;
cdfplot;
run;
The following statements create a cdf plot for width:

var length width;
cdfplot width;
run;
By default, the horizontal axis of a cdf plot is labeled with the variable name. If you specify a label for
a variable, however, the label is used. The default vertical axis label is Cumulative Percent, and the
axis is scaled in percent of observations.
If you specify a SPEC statement or a SPEC= data set in addition to the CDFPLOT statement, then the
specification limits for each variable are displayed as reference lines and are identified in a legend.
options
add features to plots. All options appear after the slash (/) in the CDFPLOT statement. In the following
example, the NORMAL option superimposes a normal cdf on the plot, and the CTEXT= option
specifies the color of the text.

cdfplot length / normal ctext=yellow;
run;
Summary of Options
The following tables list all options by function. The section “Dictionary of Options” on page 274 describes
each option in detail.
Distribution Options
You can use the options listed in Table 6.12 to superimpose a fitted theoretical distribution function on your
cdf plot.
Option Description
BETA(beta-options) Plots beta distribution with threshold
parameter , scale parameter , and shape
parameters ˛ and ˇ
EXPONENTIAL(exponential-options) Plots exponential distribution with threshold
parameter and scale parameter
GAMMA(gamma-options) Plots gamma distribution with threshold
parameter ˛
GUMBEL(Gumbel-options) Plots Gumbel distribution with location
IGAUSS(iGauss-options) Plots inverse Gaussian distribution with
mean and shape parameter
LOGNORMAL(lognormal-options) Plots lognormal distribution with threshold
parameter ,
NORMAL(normal-options) Plots normal distribution with mean and
standard deviation
PARETO(Pareto-options) Plots generalized Pareto distribution with
threshold parameter , scale parameter ,
and shape parameter ˛
POWER(power-options) Plots power function distribution with
RAYLEIGH(Rayleigh-options) Plots Rayleigh distribution with threshold
WEIBULL(Weibull-options) Plots Weibull distribution function with
and shape parameter c
Table 6.13 summarizes options that specify distribution parameters and control the display of the theoretical
distribution curve. You can specify these options in parentheses after the distribution option. For example,
the following statements use the NORMAL option to superimpose a normal distribution:
proc capability;
cdfplot / normal(mu=10 sigma=0.5 color=red);
run;
The COLOR= option specifies the color for the curve, and the normal-options MU= and SIGMA= specify
the parameters D 10 and D 0:5 for the distribution function. If you do not specify these parameters,
maximum likelihood estimates are computed.
Table 6.13 Distribution Options
Option Description
Options Used with All Distributions
COLOR= Specifies color of theoretical distribution function
L= Specifies line type of theoretical distribution function
SYMBOL= Specifies character used to plot theoretical distribution function on
line printer plots
W= Specifies width of theoretical distribution function
Beta-Options
ALPHA= Specifies first shape parameter ˛ for beta distribution function
BETA= Specifies second shape parameter ˇ for beta distribution function
SIGMA= Specifies scale parameter for beta distribution function
THETA= Specifies lower threshold parameter for beta distribution function
Exponential-Options
SIGMA= Specifies scale parameter for exponential distribution function
THETA= Specifies threshold parameter for exponential distribution
function
Gamma-Options
ALPHA= Specifies shape parameter ˛ for gamma distribution function
ALPHADELTA= Specifies change in successive estimates of ˛ at which the
Newton-Raphson approximation of ˛O terminates
ALPHAINITIAL= Specifies initial value for ˛ in the Newton-Raphson approximation
of ˛O
MAXITER= Specifies maximum number of iterations in the Newton-Raphson
approximation of ˛O
SIGMA= Specifies scale parameter for gamma distribution function
THETA= Specifies threshold parameter for gamma distribution function
Gumbel-Options
MU= Specifies location parameter for Gumbel distribution function
SIGMA= Specifies scale parameter for Gumbel distribution function
IGauss-Options
LAMBDA= Specifies shape parameter for inverse Gaussian distribution
function
MU= Specifies mean for inverse Gaussian distribution function
Lognormal-Options
SIGMA= Specifies shape parameter for lognormal distribution function
THETA= Specifies threshold parameter for lognormal distribution function
ZETA= Specifies scale parameter for lognormal distribution function
Table 6.13 continued
Option Description
Normal-Options
MU= Specifies mean for normal distribution function
SIGMA= Specifies standard deviation for normal distribution function
Pareto-Options
ALPHA= Specifies shape parameter ˛ for generalized Pareto distribution
function
SIGMA= Specifies scale parameter for generalized Pareto distribution
function
THETA= Specifies threshold parameter for generalized Pareto distribution
function
Power-Options
ALPHA= Specifies shape parameter ˛ for power function distribution
SIGMA= Specifies scale parameter for power function distribution
THETA= Specifies threshold parameter for power function distribution
Rayleigh-Options
SIGMA= Specifies scale parameter for Rayleigh distribution function
THETA= Specifies threshold parameter for Rayleigh distribution function
Weibull-Options
C= Specifies shape parameter c for Weibull distribution function
CDELTA= Specifies change in successive estimates of c at which the
Newton-Raphson approximation of cO terminates
CINITIAL= Specifies initial value for c in the Newton-Raphson approximation
of cO
MAXITER= Specifies maximum number of iterations in the Newton-Raphson
approximation of cO
SIGMA= Specifies scale parameter for Weibull distribution function
THETA= Specifies threshold parameter for Weibull distribution function
General Options
Table 6.14 General CDFPLOT Statement Options
Option Description
General Plot Layout Options
CONTENTS= Specifies table of contents entry for cdf plot grouping
HREF= Specifies reference lines perpendicular to the horizontal axis
HREFLABELS= Specifies labels for HREF= lines
NOCDFLEGEND Suppresses legend for superimposed theoretical cdf
NOECDF Suppresses plot of empirical (observed) distribution function
Option Description
NOFRAME Suppresses frame around plotting area
NOLEGEND Suppresses legend
NOSPECLEGEND Suppresses specifications legend
VREF= Specifies reference lines perpendicular to the vertical axis
VREFLABELS= Specifies labels for VREF= lines
VSCALE= Specifies scale for vertical axis
Graphics Options
ANNOTATE= Specifies annotate data set
CAXIS= Specifies color for axis
CFRAME= Specifies color for frame
CHREF= Specifies colors for HREF= lines
CSTATREF= Specifies colors for STATREF= lines
CTEXT= Specifies color for text
CVREF= Specifies colors for VREF= lines
DESCRIPTION= Specifies description for graphics catalog member
FONT= Specifies text font
HAXIS= Specifies AXIS statement for horizontal axis
HEIGHT= Specifies height of text used outside framed areas
HMINOR= Specifies number of horizontal axis minor tick marks
HREFLABPOS= Specifies position for HREF= line labels
INFONT= Specifies software font for text inside framed areas
INHEIGHT= Specifies height of text inside framed areas
LHREF= Specifies line styles for HREF= lines
LSTATREF= Specifies line styles for STATREF= lines
LVREF= Specifies line styles for VREF= lines
NAME= Specifies name for plot in graphics catalog
NOHLABEL Suppresses label for horizontal axis
NOVLABEL Suppresses label for vertical axis
NOVTICK Suppresses tick marks and tick mark labels for vertical axis
STATREF= Specifies reference lines at values of summary statistics
STATREFLABELS= Specifies labels for STATREF= lines
STATREFSUBCHAR= Specifies substitution character for displaying statistic values in
STATREFLABELS= labels
TURNVLABELS Turns and vertically strings out characters in labels for vertical axis
VAXIS= Specifies AXIS statement for vertical axis
VAXISLABEL= Specifies label for vertical axis
VMINOR= Specifies number of vertical axis minor tick marks
VREFLABPOS= Specifies position for VREF= line labels
WAXIS= Specifies line thickness for axes and frame
Options for ODS Graphics Output

ODSFOOTNOTE= Specifies footnote displayed on cdf plot
ODSFOOTNOTE2= Specifies secondary footnote displayed on cdf plot
Option Description
ODSTITLE= Specifies title displayed on cdf plot
ODSTITLE2= Specifies secondary title displayed on cdf plot
Options for Comparative Plots

ANNOKEY Applies annotation requested in ANNOTATE= data set to key cell
only
CFRAMESIDE= Specifies color for filling row label frames
CFRAMETOP= Specifies color for filling column label frames
CPROP= Specifies color for proportion of frequency bar
CTEXTSIDE= Specifies color for row labels
CTEXTTOP= Specifies color for column labels
INTERTILE= Specifies distance between tiles in comparative plot
NCOLS= Specifies number of columns in comparative plot
NROWS= Specifies number of rows in comparative plot
OVERLAY Overlays plots for different class levels (ODS Graphics only)
Options for Line Printer Charts

CDFSYMBOL= Specifies character for plotted points
HREFCHAR= Specifies line character for HREF= lines
VREFCHAR= Specifies line character for VREF= lines
The following entries provide detailed descriptions of the options specific to the CDFPLOT statement. See
“Dictionary of Common Options: CAPABILITY Procedure” on page 550 for detailed descriptions of options
common to all the plot statements.
ALPHA=value
specifies the shape parameter ˛ for distribution functions requested with the BETA, GAMMA,
PARETO, and POWER options. Enclose the ALPHA= option in parentheses after the distribu-
tion keyword. If you do not specify a value for ˛, the procedure calculates a maximum likelihood
estimate. For examples, see the entries for the distribution options.
BETA< (beta-options) >
displays a fitted beta distribution function on the cdf plot. The equation of the fitted cdf is
8
< 0 for x
F .x/ D I x .˛; ˇ/ for < x < C
:
1 for x C
where Iy .˛; ˇ/ is the incomplete beta function, and
D lower threshold parameter (lower endpoint)

D scale parameter . > 0/
˛ D shape parameter .˛ > 0/

ˇ D shape parameter .ˇ > 0/
The beta distribution is bounded below by the parameter and above by the value C . You can
specify and by using the THETA= and SIGMA= beta-options, as illustrated in the following
statements, which fit a beta distribution bounded between 50 and 75. The default values for and
are 0 and 1, respectively.
proc capability;
cdfplot / beta(theta=50 sigma=25);
run;
The beta distribution has two shape parameters, ˛ and ˇ. If these parameters are known, you can
specify their values with the ALPHA= and BETA= beta-options. If you do not specify values for ˛
and ˇ, the procedure calculates maximum likelihood estimates.
The BETA option can appear only once in a CDFPLOT statement. See Table 6.13 for a list of secondary
options you can specify with the BETA distribution option.
BETA=value
B=value
specifies the second shape parameter ˇ for beta distribution functions requested by the BETA option.
Enclose the BETA= option in parentheses after the BETA keyword. If you do not specify a value for ˇ,
the procedure calculates a maximum likelihood estimate. For examples, see the preceding entry for the
BETA option.
C=value
specifies the shape parameter c for Weibull distribution functions requested with the WEIBULL option.
Enclose the C= option in parentheses after the WEIBULL keyword. If you do not specify a value for c,
the procedure calculates a maximum likelihood estimate. You can specify the SHAPE= option as an
alias for the C= option.
CDFSYMBOL=‘character ’
specifies the character used to plot the points on legacy line printer cdf plots. The default is the plus sign
(+). This option is ignored unless you specify the LINEPRINTER option in the PROC CAPABILITY
statement. Use the SYMBOL statement to control the plotting symbol in traditional graphics output.
EXPONENTIAL< (exponential-options) >

EXP< (exponential-options ) >
displays a fitted exponential distribution function on the cdf plot. The equation of the fitted cdf is
(
F .x/ D
0 for x
x
1 exp for x >
where
D threshold parameter
The parameter must be less than or equal to the minimum data value. You can specify with the
THETA= exponential-option. The default value for is 0. You can specify with the SIGMA=
exponential-option. By default, a maximum likelihood estimate is computed for . For example, the
following statements fit an exponential distribution with D 10 and a maximum likelihood estimate
for :
proc capability;
cdfplot / exponential(theta=10 l=2 color=green);
run;
The exponential curve is green and has a line type of 2.

The EXPONENTIAL option can appear only once in a CDFPLOT statement. See Table 6.13 for a list
of secondary options you can specify with the EXPONENTIAL option.
GAMMA< (gamma-options) >

displays a fitted gamma distribution function on the cdf plot. The equation of the fitted cdf is
(
0 for x
F .x/ D 1
R x

t
˛ 1
t

.˛/ exp dt for x >
where
The parameter for the gamma distribution must be less than the minimum data value. You can specify
with the THETA= gamma-option. The default value for is 0. In addition, the gamma distribution
has a shape parameter ˛ and a scale parameter . You can specify these parameters with the ALPHA=
and SIGMA= gamma-options. By default, maximum likelihood estimates are computed for ˛ and .
For example, the following statements fit a gamma distribution function with D 4 and maximum
likelihood estimates for ˛ and :
proc capability;
cdfplot / gamma(theta=4);
run;
Note that the maximum likelihood estimate of ˛ is calculated iteratively using the Newton-Raphson
approximation. The gamma-options ALPHADELTA=, ALPHAINITIAL=, and MAXITER= control
the approximation.
The GAMMA option can appear only once in a CDFPLOT statement. See Table 6.13 for a list of
secondary options you can specify with the GAMMA option.
GUMBEL< (Gumbel-options) >

displays a fitted Gumbel distribution (also known as Type 1 extreme value distribution) function on the
cdf plot. The equation of the fitted cdf is

F .x/ D exp e .x /=
where
D location parameter
You can specify known values for and with the MU= and SIGMA= Gumbel-options. By default,
maximum likelihood estimates are computed for and .
The GUMBEL option can appear only once in a CDFPLOT statement. See Table 6.13 for a list of
secondary options you can specify with the GUMBEL option.
IGAUSS< (iGauss-options) >

displays a fitted inverse Gaussian distribution function on the cdf plot. The equation of the fitted cdf is
(r ) ( r )
x x
F .x/ D ˆ 1 C e 2= ˆ C1
x x
where ˆ./ is the standard normal cumulative distribution function, and
D mean parameter . > 0/

D shape parameter . > 0/
You can specify known values for and with the MU= and LAMBDA= iGauss-options. By default,
maximum likelihood estimates are computed for and .
The IGAUSS option can appear only once in a CDFPLOT statement. See Table 6.13 for a list of
secondary options you can specify with the IGAUSS option.
LAMBDA=value
specifies the shape parameter for distribution functions requested with the IGAUSS option. Enclose
the LAMBDA= option in parentheses after the IGAUSS distribution keyword. If you do not specify a
value for , the procedure calculates a maximum likelihood estimate.
LEGEND=name | NONE
specifies the name of a LEGEND statement describing the legend for specification limit reference lines
and superimposed distribution functions. Specifying LEGEND=NONE, which suppresses all legend
information, is equivalent to specifying the NOLEGEND option. This option is ignored unless you are
producing traditional graphics.
LOGNORMAL< (lognormal-options) >

displays a fitted lognormal distribution function on the cdf plot. The equation of the fitted cdf is
(
F .x/ D
0 for x
log.x /
ˆ for x >
D scale parameter
The parameter for the lognormal distribution must be less than the minimum data value. You can
specify with the THETA= lognormal-option. The default value for is 0. In addition, the lognormal
distribution has a shape parameter and a scale parameter . You can specify these parameters with the
SIGMA= and ZETA= lognormal-options. By default, estimates of and are computed as described
in “Lognormal Distribution” on page 353. For example, the following statements fit a lognormal
distribution function with D 10 and estimates for and :
proc capability;
cdfplot / lognormal(theta = 10);
run;
The LOGNORMAL option can appear only once in a CDFPLOT statement. See Table 6.13 for a list
of secondary options you can specify with the LOGNORMAL option.
MU=value
specifies the parameter for distribution functions requested with the GUMBEL, IGAUSS, and
NORMAL options. Enclose the MU= option in parentheses after the distribution keyword. For the
normal and inverse Gaussian distributions, the default value of is the sample mean. If you do not
specify a value for for the Gumbel distribution, the procedure calculates a maximum likelihood
estimate.
NOCDFLEGEND
suppresses the legend for the superimposed theoretical cumulative distribution function.
NOECDF
suppresses the observed distribution function (the empirical cumulative distribution function) of the
variable, which is drawn by default. This option enables you to create theoretical cdf plots without
displaying the data distribution. The NOECDF option can be used only with a theoretical distribution
(such as the NORMAL option).
NOLEGEND
suppresses legends for specification limits, theoretical distribution functions, and hidden observations.
Specifying the NOLEGEND option is equivalent to specifying LEGEND=NONE.
NORMAL< (normal-options) >

displays a fitted normal distribution function on the cdf plot. The equation of the fitted cdf is
F .x/ D ˆ x for 1 < x < 1

D mean
D standard deviation . > 0/
You can specify known values for and with the MU= and SIGMA= normal-options, as shown in
the following statements:
proc capability;
cdfplot / normal(mu=14 sigma=.05);
run;
By default, the sample mean and sample standard deviation are calculated for and . The NORMAL
option can appear only once in a CDFPLOT statement. For an example, see Output 6.4.1. See
Table 6.13 for a list of secondary options you can specify with the NORMAL option.
NOSPECLEGEND
NOSPECL
suppresses the portion of the legend for specification limit reference lines.
PARETO< (Pareto-options) >

displays a fitted generalized Pareto distribution function on the cdf plot. The equation of the fitted cdf
is
˛1
˛.x /
F .x/ D 1 1

where
˛ D shape parameter
The parameter for the generalized Pareto distribution must be less than the minimum data value.
You can specify with the THETA= Pareto-option. The default value for is 0. In addition, the
generalized Pareto distribution has a shape parameter ˛ and a scale parameter . You can specify
these parameters with the ALPHA= and SIGMA= Pareto-options. By default, maximum likelihood
estimates are computed for ˛ and .
The PARETO option can appear only once in a CDFPLOT statement. See Table 6.13 for a list of
secondary options you can specify with the PARETO option.
POWER< (power-options) >

displays a fitted power function distribution on the cdf plot. The equation of the fitted cdf is
8
< 0
ˆ ˛ for x
x
F .x/ D for < x < C
ˆ
1 for x C
:
where

The power function distribution is bounded below by the parameter and above by the value C .
You can specify and by using the THETA= and SIGMA= power-options. The default values for
and are 0 and 1, respectively.
You can specify a value for the shape parameter, ˛, with the ALPHA= power-option. If you do not
specify a value for ˛, the procedure calculates a maximum likelihood estimate.
The power function distribution is a special case of the beta distribution with its second shape parameter,
ˇ D 1.
The POWER option can appear only once in a CDFPLOT statement. See Table 6.13 for a list of
secondary options you can specify with the POWER option.
RAYLEIGH< (Rayleigh-options) >

displays a fitted Rayleigh distribution function on the cdf plot. The equation of the fitted cdf is
.x /2 =.2 2 /
F .x/ D 1 e
where
The parameter for the Rayleigh distribution must be less than the minimum data value. You can
specify with the THETA= Rayleigh-option. The default value for is 0. You can specify with the
SIGMA= Rayleigh-option. By default, a maximum likelihood estimate is computed for .
The RAYLEIGH option can appear only once in a CDFPLOT statement. See Table 6.13 for a list of
secondary options you can specify with the RAYLEIGH option.
SIGMA=value
specifies the parameter for distribution functions requested by the BETA, EXPONENTIAL,
GAMMA, GUMBEL, LOGNORMAL, NORMAL, PARETO, POWER, RAYLEIGH, and WEIBULL
options. Enclose the SIGMA= option in parentheses after the distribution keyword. Table 6.15
summarizes the use of the SIGMA= option.
Table 6.15 SIGMA= Option Summary
Distribution Option SIGMA= Specifies Default Value Alias

BETA Scale parameter 1 SCALE=
POWER
EXPONENTIAL Scale parameter Maximum likelihood estimate SCALE=
GAMMA
WEIBULL
GUMBEL Scale parameter Maximum likelihood estimate
PARETO
RAYLEIGH
LOGNORMAL Shape parameter Estimate computed as described SHAPE=
in “Lognormal Distribution” on
page 353
NORMAL Scale parameter Standard deviation
SYMBOL=‘character ’
specifies the character used to plot the theoretical distribution function on legacy line printer plots.
Enclose the SYMBOL= option in parentheses after the distribution option. The default character
is the first letter of the distribution option keyword. This option is ignored unless you specify the
LINEPRINTER option in the PROC CAPABILITY statement.
THETA=value
THRESHOLD=value
specifies the lower threshold parameter for theoretical cumulative distribution functions requested
with the BETA, EXPONENTIAL, GAMMA, LOGNORMAL, PARETO, POWER, RAYLEIGH, and
WEIBULL options. Enclose the THETA= option in parentheses after the distribution keyword. The
default value is 0.
VSCALE=PERCENT | PROPORTION
specifies the scale of the vertical axis. The value PERCENT scales the data in units of percent
of observations per data unit. The value PROPORTION scales the data in units of proportion of
observations per data unit. The default is PERCENT.
WEIBULL< (Weibull-options) >

displays a fitted Weibull distribution function on the cdf plot. The equation of the fitted cdf is
(
F .x/ D
0 c for x
x
1 exp for x >
where
c D shape parameter .c > 0/
The parameter must be less than the minimum data value. You can specify with the THETA=
Weibull-option. The default value for is 0. In addition, the Weibull distribution has a shape parameter
c and a scale parameter . You can specify these parameters with the SIGMA= and C= Weibull-options.
By default, maximum likelihood estimates are computed for c and . For example, the following
statements fit a Weibull distribution function with D 15 and maximum likelihood estimates for
and c:
proc capability;
cdfplot / weibull(theta=15);
run;
Note that the maximum likelihood estimate of c is calculated iteratively using the Newton-Raphson
approximation. The Weibull-options CDELTA=, CINITIAL=, and MAXITER= control the approxima-
tion.
The WEIBULL option can appear only once in a CDFPLOT statement. See Table 6.13 for a list of
secondary options you can specify with the WEIBULL option.
ZETA=value
specifies a value for the scale parameter for a lognormal distribution function requested with the
LOGNORMAL option. Enclose the ZETA= option in parentheses after the LOGNORMAL keyword.
If you do not specify a value for , a maximum likelihood estimate is computed. You can specify the
SCALE= option as an alias for the ZETA= option.
Details: CDFPLOT Statement

ODS Graphics
Before you create ODS Graphics output, ODS Graphics must be enabled (for example, by using the ODS
GRAPHICS ON statement). For more information about enabling and disabling ODS Graphics, see the
section “Enabling and Disabling ODS Graphics” (Chapter 24, SAS/STAT User’s Guide).
The appearance of a graph produced with ODS Graphics is determined by the style associated with the ODS
destination where the graph is produced. CDFPLOT options used to control the appearance of traditional
graphics are ignored for ODS Graphics output.
When ODS Graphics is in effect, the CDFPLOT statement assigns a name to the graph it creates. You can
use this name to reference the graph when using ODS. The name is listed in Table 6.16.
Table 6.16 ODS Graphics Produced by the CDFPLOT Statement
ODS Graph Name Plot Description

CDFPlot Cumulative distribution function plot
See Chapter 4, “SAS/QC Graphics,” for more information about ODS Graphics and other methods for
producing charts.
Examples: CDFPLOT Statement

This section illustrates how to display a fitted distribution function, inset tables, and display reference lines
on your cdf plot.
Example 6.4: Fitting a Normal Distribution

You can use the CDFPLOT statement to fit any of eleven theoretical distributions (beta, exponential, gamma,
Gumbel, inverse Gaussian, lognormal, normal, generalized Pareto, power function, Rayleigh, and Weibull)
and superimpose them on the cdf plot. The following statements use the NORMAL option to display a fitted
normal distribution function on a cdf plot of breaking strengths. The data set Cord is given in Figure 6.5, and
the plot is shown in Output 6.4.1.
title 'Cumulative Distribution Function of Breaking Strength';

spec lsl=6.8;
cdf Strength / normal
odstitle=title;
inset mean std pctlss / format = 5.2
header = "Summary Statistics";
run;
Examples: CDFPLOT Statement F 283
Output 6.4.1 Superimposed Normal Distribution Function
The NORMAL option requests the fitted curve. The INSET statement requests an inset containing the
mean, the standard deviation, and the percent of observations below the lower specification limit. For more
information about the INSET statement, see “INSET Statement: CAPABILITY Procedure” on page 398. The
SPEC statement requests a lower specification limit at 6.8. For more information about the SPEC statement,
see “SPEC Statement” on page 225.
The agreement between the empirical and the normal distribution functions in Output 6.4.1 is evidence that
the normal distribution is an appropriate model for the distribution of breaking strengths.
The CAPABILITY procedure provides a variety of other tools for assessing goodness of fit. Goodness-of-
fit tests (see “Printed Output” on page 362) provide a quantitative assessment of a proposed distribution.
Probability and Q-Q plots, created with the PROBPLOT (“PROBPLOT Statement: CAPABILITY Procedure”
on page 477), QQPLOT (“QQPLOT Statement: CAPABILITY Procedure” on page 508), and PPPLOT
(“PPPLOT Statement: CAPABILITY Procedure” on page 454) statements, provide effective graphical
diagnostics.
Example 6.5: Using Reference Lines with CDF Plots

Customer requirements dictate that the breaking strengths in the previous example have upper and lower
specification limits of 7.2 and 6.8 psi, respectively. Moreover, less than 5% of the cords can have breaking
strengths outside the limits.
The following statements create a cdf plot with reference lines at the 5% and 95% cumulative percent levels:

spec lsl=6.8 usl=7.2;
cdf Strength / vref = 5 95
vreflabels = '5%' '95%'
odstitle = title;
inset pctgtr pctlss / format = 5.2
pos = e
header = "Summary Statistics";
run;
The INSET statement requests an inset with the percentages of measurements above the upper limit and below
the lower limit. For more information about the INSET statement, see “INSET Statement: CAPABILITY
Procedure” on page 398.
In Output 6.5.1, the empirical cdf is below the intersection between the lower specification limit line and the
5% line, so less than 5% of the measurements are below the lower limit. The ecdf, however, is also below the
intersection between the upper specification limit line and the 95% line, implying that more than 5% of the
measurements are greater than the upper limit. Thus, the goal of having less than 5% of the measurements
above the upper specification limit has not been met.
COMPHISTOGRAM Statement: CAPABILITY Procedure F 285
Output 6.5.1 Reference Lines with a Cumulative Distribution Function Plot
COMPHISTOGRAM Statement: CAPABILITY Procedure
Overview: COMPHISTOGRAM Statement

Comparative histograms are useful for comparing the distribution of a process variable across levels of
classification variables. You can use the COMPHISTOGRAM statement to create one-way and two-way
comparative histograms. When used with a single classification variable, the COMPHISTOGRAM statement
displays an array of component histograms (stacked or side-by-side), one for each level of the classification
variable. When used with two classification variables, the COMPHISTOGRAM statement displays a matrix
of component histograms, one for each combination of levels of the classification variables.
In quality improvement applications, typical uses of comparative histograms include
comparing the capability of a process before and after an improvement

comparing process capabilities of two or more suppliers
exploring stratification in process data due to different lots, machines, manufacturing methods, and so
forth
studying the evolution of process capability over successive time periods
You can use options in the COMPHISTOGRAM statement to
specify the midpoints or endpoints for histogram intervals
specify the number of rows and/or columns of component histograms
display specification limits on the component histograms
display density curves for fitted normal distributions
display kernel density estimates
request graphical enhancements
inset summary statistics and process capability indices on the component histograms
You have two alternatives for producing comparative histograms with the COMPHISTOGRAM statement:
Otherwise, traditional graphics are produced if SAS/GRAPH is licensed.
N OTE : You cannot use the COMPHISTOGRAM statement together with the CLASS statement.
Getting Started: COMPHISTOGRAM Statement

This section introduces the COMPHISTOGRAM statement with examples that illustrate commonly used
options. Complete syntax for the COMPHISTOGRAM statement is presented in the section “Syntax:
COMPHISTOGRAM Statement” on page 290, and advanced examples are given in the section “Examples:
COMPHISTOGRAM Statement” on page 306.
Creating a One-Way Comparative Histogram

N OTE : See Comparative Histograms with Normal Curves in the SAS/QC Sample Library.
The effective channel length (in microns) is measured for 1225 field effect transistors. The channel lengths
are saved as values of the variable Length in a SAS data set named Channel:
Getting Started: COMPHISTOGRAM Statement F 287
data Channel;
length Lot $ 16;
input Length @@;
select;
when (_n_ <= 425) Lot='Lot 1';
when (_n_ >= 926) Lot='Lot 3';
otherwise Lot='Lot 2';
end;
datalines;
0.91 1.01 0.95 1.13 1.12 0.86 0.96 1.17 1.36 1.10
0.98 1.27 1.13 0.92 1.15 1.26 1.14 0.88 1.03 1.00
0.98 0.94 1.09 0.92 1.10 0.95 1.05 1.05 1.11 1.15
1.11 0.98 0.78 1.09 0.94 1.05 0.89 1.16 0.88 1.19
1.01 1.08 1.19 0.94 0.92 1.27 0.90 0.88 1.38 1.02
... more lines ...
2.13 2.05 1.90 2.07 2.15 1.96 2.15 1.89 2.15 2.04
1.95 1.93 2.22 1.74 1.91
;
The data set Channel is also used in Example 6.12, where a kernel density estimate is superimposed on the
histogram of channel lengths. The display in Output 6.12.1 reveals that there are three distinct peaks in the
process distribution. To investigate whether these peaks (modes) in the histogram are related to the lot source,
you can create a comparative histogram that uses Lot as a classification variable. The following statements
create the comparative histogram shown in Figure 6.6:
title "Comparative Analysis of Lot Source";

proc capability data=Channel noprint;
specs lsl = 0.8 usl = 2.0;
comphist Length / class = Lot
nrows = 3
nlegend = 'Lot Size'
nlegendpos = nw
odstitle = title;
label Lot = 'Transistor Source';
run;
The COMPHISTOGRAM statement requests a comparative histogram for the process variable Length. The
CLASS= option requests a component histogram for each level (distinct value) of the classification variable
Lot. The option NROWS=3 stacks the histograms three to a page. The NLEGEND= option adds a sample
size legend to each component histogram, and the option NLEGENDPOS=NW positions each legend in the
northwest corner. The SPEC statement provides the specification limits displayed as vertical reference lines.
See the section “Dictionary of Options” on page 295 for descriptions of these options, and see the section
“SPEC Statement” on page 225 for details of the SPEC statement.
Figure 6.6 Comparison by Lot Source
Adding Fitted Normal Curves to a Comparative Histogram

N OTE : See Comparative Histograms with Normal Curves in the SAS/QC Sample Library.
In Figure 6.6, it appears that each lot produces transistors with channel lengths that are normally distributed.
The following statements use the NORMAL option to fit a normal distribution to the data for each lot (the
observations corresponding to a specific level of the classification variable are referred to as a cell). The
normal parameters and are estimated from the data for each lot, and the curves are superimposed on
each component histogram.
title "Comparative Analysis of Lot Source";

comphist Length / class = Lot
nrows = 3
intertile = 1
odstitle = title
cprop
normal;
label Lot = 'Transistor Source';
Getting Started: COMPHISTOGRAM Statement F 289
run;
The comparative histogram is displayed in Figure 6.7.
Figure 6.7 Fitting Normal Curves
Specifying INTERTILE=1 inserts a space of one percent screen unit between the framed areas, which are
referred to as tiles. The shaded bars, added with the CPROP= option, represent the relative frequency of
observations in each cell. See “Dictionary of Options” on page 295 for details concerning these options.
Syntax: COMPHISTOGRAM Statement

The syntax for the COMPHISTOGRAM statement is as follows:
COMPHISTOGRAM < variables > / CLASS=(class-variables) < options > ;
You can specify the keyword COMPHIST as an alias for COMPHISTOGRAM. You can use any number of
COMPHISTOGRAM statements after a PROC CAPABILITY statement.
To create a comparative histogram, you must specify at least one variable and either one or two class-variables
(also referred to as classification variables). The COMPHISTOGRAM statement displays a component
histogram of the values of the variable for each level of the class-variables. The observations in a particular
level are referred to as a cell.
The components of the COMPHISTOGRAM statement are described as follows:
variables
are the process variables for which comparative histograms are to be created. If you specify a VAR
statement, the variables must also be listed in the VAR statement. Otherwise, variables can be any
numeric variables in the input data set that are not also listed as class-variables. If you do not specify
variables in a COMPHISTOGRAM statement or a VAR statement, then by default a comparative
histogram is created for each numeric variable in the DATA= data set that is not used as a class-variable.
If you use a VAR statement and do not specify variables in the COMPHISTOGRAM statement, then
by default a comparative histogram is created for each variable listed in the VAR statement.
For example, suppose a data set named steel contains two process variables named length and width,
a numeric classification variable named lot, and a character classification variable named day. The
following statements create two comparative histograms, one for length and one for Width:

comphist / class = lot;
run;
Likewise, the following statements create comparative histograms for length and width:

var length width;
comphist / class = day;
run;
The following statements create three comparative histograms (for length, width, and lot):

comphist / class = day;
run;
The following statements create a comparative histogram for Width only:

Syntax: COMPHISTOGRAM Statement F 291

var length width;
comphist width / class=lot;
run;
class-variables
are one or two required classification variables. For example, the following statements create a one-way
comparative histogram for width by using the classification variable lot:

comphist width / class=lot;
run;
The following statements create a two-way comparative histogram for width classified by lot and day:

comphist width / class=(lot day);
run;
Note that the parentheses surrounding the class-variables are needed only if two classification variables
are specified. See Output 6.6.1 and Output 6.7.1 for further examples.
options
control the features of the comparative histogram. All options are specified after the slash (/) in the
COMPHIST statement. In the following example, the CLASS= option specifies the classification
variable, the NORMAL option fits a normal density curve in each cell, and the CTEXT= option
specifies the color of the text:

comphist length / class = lot
normal
ctext = yellow;
run;
Summary of Options
The following tables list the COMPHIST statement options by function. For complete descriptions, see
“Dictionary of Options” on page 295.
Table 6.17 lists the options for requesting that a fitted normal distribution or a kernel density estimate be
overlaid on the comparative histogram.
Table 6.17 Density Estimation Options
Option Description
KERNEL(kernel-options) Fits kernel density estimates
NORMAL(normal-options) Fits normal distribution with mean and
standard deviation
You can specify the secondary options listed in Table 6.18 in parentheses after the KERNEL option to control
features of kernel density estimates.
Option Description
C= Specifies standardized bandwidth parameter c for kernel
density estimate
COLOR= Specifies color of the kernel density curve
FILL Fills area under kernel density curve
K= Specifies NORMAL, TRIANGULAR, or QUADRATIC
kernel
L= Specifies line type used for kernel density curve
LOWER= Specifies lower bound for kernel density curve
UPPER= Specifies upper bound for kernel density curve
W= Specifies line width for kernel density curve
You can specify the secondary options listed in Table 6.19 in parentheses after the NORMAL option to
control features of fitted normal distributions.
Table 6.19 Normal-Options
Option Description
COLOR= Specifies color of normal curve
FILL Fills area under normal curve
L= Specifies line type of normal curve
MU= Specifies mean for fitted normal curve
SIGMA= Specifies standard deviation for fitted normal curve
W= Specifies width of normal curve
For example, the following statements use the NORMAL option to fit a normal curve in each cell of the
comparative histogram:
proc capability;
comphistogram / class = machine
normal(color=red l=2);
run;
The COLOR= normal-option draws the curve in red, and the L= normal-option specifies a line style of 2
(a dashed line) for the curve. In this example, maximum likelihood estimates are computed for the normal
parameters and for each cell because these parameters are not specified.
General Options
Table 6.20 General COMPHISTOGRAM Statement Options
Option Description
Classification Options
CLASS= Specifies classification variables
CLASSKEY= Specifies key cell
MISSING1 Requests that missing values of first CLASS= variable be treated
as a level of that CLASS= variable
MISSING2 Requests that missing values of second CLASS= variable be
treated as a level of that CLASS= variable
ORDER1= Specifies display order for values of the first CLASS= variable
ORDER2= Specifies display order for values of the second CLASS= variable
Layout Options
BARLABEL= Produces labels above histogram bars
BARWIDTH= Specifies width for the bars
CLIPSPEC= Clips histogram bars at specification limits if there are no
observations beyond the limits
ENDPOINTS= Labels interval endpoints and specifies how they are determined
HOFFSET= Specifies offset for horizontal axis
INTERTILE= Specifies distance between tiles
MAXNBIN= Specifies maximum number of bins displayed
MAXSIGMAS= Limits number of bins displayed to range of value standard
deviations above and below mean of data in key cell
MIDPOINTS= Specifies how midpoints are determined
NCOLS= Specifies number of columns in comparative histogram
NOBARS Suppresses histogram bars
NOKEYMOVE Suppresses rearrangement of cells that occurs by default with the
CLASSKEY= option
NOPLOT Suppresses plot
NROWS= Specifies number of rows in comparative histogram
RTINCLUDE Includes right endpoint in interval
WBARLINE= Specifies line thickness for bar outlines
Axis and Legend Options

GRID Adds grid corresponding to vertical axis
LGRID= Specifies line style for grid requested with GRID option
NLEGEND Specifies form of the legend displayed inside tiles
NLEGENDPOS= Specifies position of legend displayed inside tiles
Option Description
TILELEGLABEL= Specifies label displayed when _CTILE_ and _TILELG_ variables
are provided in the CLASSSPEC= data set
TURNVLABELS Turns and strings out vertically characters in vertical axis labels
VAXIS= Specifies tick mark values for vertical axis
VOFFSET= Specifies length of offset at upper end of vertical axis
WGRID= Specifies line thickness for grid
Reference Line Options

FRONTREF Draws reference lines in front of histogram bars
HREF= Specifies reference lines perpendicular to horizontal axis
HREFLABPOS= Specifies vertical position of labels for HREF= lines
LHREF= Specifies line style for HREF= lines
LVREF= Specifies line style for VREF= lines
VREF= Specifies reference lines perpendicular to vertical axis
VREFLABPOS= Specifies horizontal position of labels for VREF= lines
Text Enhancement Options

FONT= Specifies software font for text
Color and Pattern Options

CBARLINE= Specifies color for outline of the bars
CFILL= Specifies color for filling bars
CFRAMENLEG= Specifies the color for the frame requested by the NLEGEND
option
CFRAMESIDE= Specifies color for filling frame for row labels
CFRAMETOP= Specifies color for filling frame for column labels
CGRID= Specifies color for grid lines
CHREF= Specifies color for HREF= lines
CVREF= Specifies color for VREF= lines
Option Description
PFILL= Specifies pattern used to fill bars
Input and Output Data Set Options

ANNOKEY Applies annotation requested in ANNOTATE= data set to key cell
only
ANNOTATE= Annotate data set
CLASSSPEC= Data set with specification limit information for each cell
OUTHISTOGRAM= Information on histogram intervals
Graphics Catalog Options

DESCRIPTION= Specifies description for graphics catalog member
The following sections describe in detail the options specific to the COMPHISTOGRAM statement. See
General Options
graphics:
BARLABEL=COUNT | PERCENT | PROPORTION

displays labels above the histogram bars. If you specify BARLABEL=COUNT, the label shows the
number of observations associated with a particular bar. BARLABEL=PERCENT shows the percent
of observations represented by that bar. If you specify BARLABEL=PROPORTION, the label displays
the proportion of observations associated with the bar.
C=value-list | MISE
specifies the standardized bandwidth parameter c for kernel density estimates requested with the
KERNEL option. You can specify up to five values to display multiple estimates in each cell. You can
also specify the keyword MISE to request the bandwidth parameter that minimizes the estimated mean
integrated square error (MISE). For example, consider the following statements (for more information,
see “Kernel Density Estimates” on page 360):
proc capability;
comphist length / class=batch kernel(c = 0.5 1.0 mise);
run;
The KERNEL option displays three density estimates. The first two have standardized bandwidths
of 0.5 and 1.0, respectively. The third has a bandwidth parameter that minimizes the MISE. You can
also use the C= and K= options (K= specifies kernel type) to display multiple estimates. For example,
consider the following statements:
proc capability;
comphist length / class = batch
kernel(c = 0.75 k = normal triangular);
run;
Here two estimates are displayed. The first uses a normal kernel and bandwidth parameter of 0.75, and
the second uses a triangular kernel and a bandwidth parameter of 0.75. In general, if more kernel types
are specified than bandwidth parameters, the last bandwidth parameter in the list will be repeated for
the remaining estimates. Likewise, if more bandwidth parameters are specified than kernel types, the
last kernel type will be repeated for the remaining estimates. The default is MISE.
CLASS=variable
CLASS=(variable1 variable2)
specifies that a comparative histogram is to be created using the levels of the variables (also referred to
as class-variables or classification variables).
If you specify a single variable, a one-way comparative histogram is created. The observations in
the input data set are sorted by the formatted values (levels) of the variable. A separate histogram is
created for the process variable values in each level, and these component histograms are arranged in
an array to form the comparative histogram. Uniform horizontal and vertical axes are used to facilitate
comparisons. For an example, see Figure 6.6.
If you specify two classification variables, a two-way comparative histogram is created. The obser-
vations in the input data set are cross-classified according to the values (levels) of these variables. A
separate histogram is created for the process variable values in each cell of the cross-classification, and
these component histograms are arranged in a matrix to form the comparative histogram. The levels
of variable1 are used to label the rows of the matrix, and the levels of variable2 are used to label the
columns of the matrix. Uniform horizontal and vertical axes are used to facilitate comparisons. For an
example, see Output 6.7.1.
Classification variables can be numeric or character. Formatted values are used to determine the levels.
You can specify whether missing values are to be treated as a level with the MISSING1 and MISSING2
options.
If a label is associated with a classification variable, the label is displayed on the comparative histogram.
The variable label is displayed parallel to the column (or row) labels. For an example, see Figure 6.6.
CLASSKEY=‘value’
CLASSKEY=(‘value1’ ‘value2’)
specifies the key cell in a comparative histogram requested with the CLASS= option. The bin size and
midpoints are first determined for the key cell, and then the midpoint list is extended to accommodate
the data ranges for the remaining cells. Thus, the choice of the key cell determines the uniform
horizontal axis used for all cells.
If you specify CLASS=variable, you can specify CLASSKEY=’value’ to identify the key cell as the
level for which variable is equal to value. You must specify a formatted value. By default, the levels
are sorted in the order determined by the ORDER1= option, and the key cell is the level that occurs
first in this order. The cells are displayed in this order from top to bottom (or left to right), and,
consequently, the key cell is displayed at the top or at the left. If you specify a different key cell with
the CLASSKEY= option, this cell is displayed at the top or at the left unless you also specify the
NOKEYMOVE option.
If you specify CLASS=(variable1 variable2), you can specify CLASSKEY=(’value1’ ’value2’) to
identify the key cell as the level for which variable1 is equal to value1 and variable2 is equal to
value2. Here, value1 and value2 must be formatted values, and they must be enclosed in quotes. For
an example of the CLASSKEY= option with a two-way comparative histogram, see Output 6.7.1. By
default, the levels of variable1 are sorted in the order determined by the ORDER1= option, and within
each of these levels, the levels of variable2 are sorted in the order determined by the ORDER2= option.
The default key cell is the combination of levels of variable1 and variable2 that occurs first in this
order. The cells are displayed in order of variable1 from top to bottom and in order of variable2 from
left to right. Consequently, the default key cell is displayed in the upper left corner. If you specify a
different key cell with the CLASSKEY= option, this cell is displayed in the upper left corner unless
you also specify the NOKEYMOVE option.
CLASSSPEC=SAS-data-set
CLASSSPECS=SAS-data-set
specifies a data set that provides distinct specification limits for each cell, as well as a color, legend,
and label for the corresponding tile. Table 6.21 lists the variables that are read from a CLASSSPEC=
data set.
Table 6.21 Variables Read from CLASSSPEC= Data Set
Variable Name Description

BY variables Subsets the data set
Classification variables Specifies the structure of the comparative histogram
_VAR_ Specifies name of process variable (must be character
variable of length 8)
_LSL_ Specifies lower specification limit for tile
_TARGET_ Specifies target value for tile
_USL_ Specifies upper specification limit for tile
_CTILE_ Specifies background color for tiles (must be character
variable of length 8)
_TILELG_ Specifies text displayed in color tile legend at bottom
of comparative histogram (character variable of length
not greater than 16)
_TILELB_ Specifies text displayed in corner of each tile (character
variable of length not greater than 16)
If you specify a CLASSSPEC= data set, you cannot use the SPEC statement or a SPEC= data set. If
you use a BY statement, the CLASSSPEC= data set must contain one observation for each unique
combination of process and classification variables within each BY group. See Example 6.6 for an
example of a CLASSSPEC= data set.
Also note that
you can suppress the background color for a tile by assigning the value ‘EMPTY’ or a blank
value to the variable _CTILE_
you can use the NLEGENDPOS= option to specify the corner of the tile in which the _TILELB_
label is displayed. You can frame the label with the CFRAMENLEG= option.
you cannot use the variable _TILELG_ unless you specify the variable _CTILE_
the variable _TILELB_ takes precedence over the NLEGEND option
ENDPOINTS=value-list | KEY | UNIFORM

specifies that histogram interval endpoints, rather than midpoints, are aligned with horizontal axis tick
marks, and specifies how the endpoints are determined. The method you specify is used for all process
variables analyzed with the COMPHISTOGRAM statement.
If you specify ENDPOINTS=value-list, the values must be listed in increasing order and must be
evenly spaced. The difference between consecutive endpoints is used as the width of the histogram
bars. The first value is the lower bound of the first histogram bin and the last value is the upper bound
of the last bin. Thus, the number of values in the list is one greater than the number of bins it specifies.
If the range of the values does not cover the range of the data as well as any specification limits (LSL
and USL) that are given, the list is extended in either direction as necessary.
If you specify ENDPOINTS=KEY, the procedure first determines the endpoints for the data in the
key cell. The initial number of endpoints is based on the number of observations in the key cell by
using the method of Terrell and Scott (1985). The endpoint list for the key cell is then extended in
either direction as necessary until it spans the data in the remaining cells. If the key cell contains no
observations, the method of determining bins reverts to ENDPOINTS=UNIFORM.
If you specify ENDPOINTS=UNIFORM, the procedure determines the endpoints by using all the
observations as if there were no cells. In other words, the number of endpoints is computed from the
total sample size by using the method of Terrell and Scott (1985).
FILL
fills areas under a fitted density curve with colors and patterns. Enclose the FILL option in parentheses
after the keyword NORMAL or KERNEL. Depending on the area to be filled (outside or between the
specification limits), you can specify the color and pattern with options in the SPEC statement and the
COMPHISTOGRAM statement, as summarized in the following table:
Area Under Curve Statement Color Option Pattern Option

Between specification limits COMPHIST CFILL=color PFILL=pattern
Left of lower specification limit SPEC CLEFT=color PLEFT=pattern
Right of upper specification limit SPEC CRIGHT=color PRIGHT=pattern
If you do not display specification limits, you can use the CFILL= and PFILL= options to specify
the color and pattern for the entire area under the curve. Solid fills are used by default if patterns
are not specified. You can specify the FILL option with only one fitted curve. For an example, see
Output 6.6.1. Refer to SAS/GRAPH: Reference for a list of available patterns and colors. If you do not
specify the FILL option but you do specify the options in the preceding table, the colors and patterns
are applied to the corresponding areas under the histogram.
GRID
adds a grid to the comparative histogram. Grid lines are horizontal lines positioned at major tick marks
on the vertical axis.
INTERTILE=value
specifies the distance in horizontal percent screen units between tiles. For an example, see Figure 6.7.
By default, the tiles are contiguous.
K=NORMAL | TRIANGULAR | QUADRATIC

specifies the type of kernel (normal, triangular, or quadratic) used to compute kernel density estimates
requested with the KERNEL option. Enclose the K= option in parentheses after the keyword KERNEL.
You can specify a single type or a list of types. If you specify more estimates than types, the last kernel
type in the list is used for the remaining estimates. By default, a normal kernel is used.
KERNEL< ( kernel-options ) >

requests a kernel density estimate for each cell of the comparative histogram. You can specify the
kernel-options described in the following table:
Option Description
FILL Specifies that the area under the curve is to be filled
COLOR= Specifies the color of the curve
L= Specifies the line style for the curve
W= Specifies the width of the curve
K= Specifies the type of kernel
C= Specifies the smoothing parameter
LOWER= Specifies the lower bound for the curve
UPPER= Specifies the upper bound for the curve
See Output 6.6.1 for an example. By default, the estimate is based on the AMISE method. For more
information, see “Kernel Density Estimates” on page 360.
LOWER=value
specifies the lower bound for a kernel density estimate curve. Enclose the LOWER= option in
parentheses after the KERNEL option. You can specify a single lower bound or a list of lower bounds.
By default, a kernel density estimate curve has no lower bound.
MAXNBIN=n
specifies the maximum number of bins to be displayed. This option is useful in situations where the
scales or ranges of the data distributions differ greatly from cell to cell. By default, the bin size and
midpoints are determined for the key cell, and then the midpoint list is extended to accommodate the
data ranges for the remaining cells. However, if the cell scales differ considerably, the resulting number
of bins may be so great that each cell histogram is scaled into a narrow region. By limiting the number
of bins with the MAXNBIN= option, you can narrow the window about the data distribution in the key
cell. Note that the MAXNBIN= option provides an alternative to the MAXSIGMAS= option.
MAXSIGMAS=value
limits the number of bins to be displayed to a range of value standard deviations (of the data in the key
cell) above and below the mean of the data in the key cell. This option is useful in situations where the
data ranges for the remaining cells. If the cell scales differ considerably, however, the resulting number
of bins with the MAXSIGMAS= option, you narrow the window about the data distribution in the key
cell. Note that the MAXSIGMAS= option provides an alternative to the MAXNBIN= option.
MIDPOINTS=value-list | KEY | UNIFORM

specifies how midpoints are determined for the bins in the comparative histogram. The method you
specify is used for all process variables analyzed with the COMPHISTOGRAM statement.
If you specify MIDPOINTS=value-list, the values must be listed in increasing order and must be
evenly spaced. The difference between consecutive midpoints is used as the width of the histogram
bars. If the range of the values does not cover the range of the data as well as any specification limits
(LSL and USL) that are given, the list is extended in either direction as necessary. See Example 6.6 for
an illustration.
If you specify MIDPOINTS=KEY, the procedure first determines the midpoints for the data in the key
cell. The initial number of midpoints is based on the number of observations in the key cell by using
the method of Terrell and Scott (1985). The midpoint list for the key cell is then extended in either
direction as necessary until it spans the data in the remaining cells.
If you specify MIDPOINTS=UNIFORM, the procedure determines the midpoints using all the obser-
vations as if there were no cells. In other words, the number of midpoints is computed from the total
sample size by using the method of Terrell and Scott (1985).
By default, MIDPOINTS=KEY. However, if the key cell contains no observations, the default is
MIDPOINTS=UNIFORM.
MISSING1
specifies that missing values of the first CLASS= variable are to be treated as a level of the CLASS=
variable. If the first CLASS= variable is a character variable, a missing value is defined as a blank
internal (unformatted) value. If the process variable is numeric, a missing value is defined as any of
the SAS System missing values. If you do not specify MISSING1, observations for which the first
CLASS= variable is missing are excluded from the analysis.
MISSING2
specifies that missing values of the second CLASS= variable are to be treated as a level of the CLASS=
variable. If the second CLASS= variable is a character variable, a missing value is defined as a blank
internal (unformatted) value. If the process variable is numeric, a missing value is defined as any of
the SAS System missing values. If you do not specify MISSING2, observations for which the second
CLASS= variable is missing are excluded from the analysis.
MU=value
specifies the parameter for the normal density curves requested with the NORMAL option. Enclose
the MU= option in parentheses after the NORMAL option. The default value is the sample mean of
the observations in the cell.
NOBARS
suppresses the display of the bars in a comparative histogram.
NOCHART
suppresses the creation of a comparative histogram. This is an alias for NOPLOT.
NOKEYMOVE
suppresses the rearrangement of cells that occurs by default when you use the CLASSKEY= option to
specify the key cell. For details, see the entry for the CLASSKEY= option.
NOPLOT
suppresses the creation of a comparative histogram. This option is useful when you are using the
COMPHISTOGRAM statement solely to create an output data set.

displays a normal density curve for each cell of the comparative histogram. The equation of the normal
density curve is
hv
exp 12 . x /2 for 1 < x < 1

p.x/ D p 2
where
D mean
h D width of histogram interval
v D vertical scaling factor
and
8
< n the sample size, for VSCALE=COUNT
vD 100 for VSCALE=PERCENT
1 for VSCALE=PROPORTION
:
If you specify values for and with the MU= and SIGMA= normal-options, the same curve is
displayed for each cell. By default, a distinct curve is displayed for each cell based on the sample mean
and standard deviation for that cell. For example, the following statements display a distinct curve for
each level of the variable Supplier:
proc capability noprint;

comphist width / class=supplier normal(color=red l=2);
run;
The curves are drawn in red with a line style of 2 (a dashed line). See Figure 6.7 for another illustration.
Table 6.19 lists options that can be specified in parentheses after the NORMAL option.
ORDER1=INTERNAL | FORMATTED | DATA | FREQ

specifies the display order for the values of the first CLASS= variable.
The levels of the first CLASS= variable are always constructed using the formatted values of the vari-
able, and the formatted values are always used to label the rows (columns) of a comparative histogram.
You can use the ORDER1= option to determine the order of the rows (columns) corresponding to these
values, as follows:
If you specify ORDER1=INTERNAL, the rows (columns) are displayed from top to bottom
(left to right) in increasing order of the internal (unformatted) values of the first CLASS= variable.
If there are two or more distinct internal values with the same formatted value, then the order is
determined by the internal value that occurs first in the input data set.
For example, suppose that you specify a numeric CLASS= variable called Day (with values 1, 2,
and 3). Suppose also that a format (created with the FORMAT procedure) is associated with Day
and that the formatted values are as follows: 1 = ’Wednesday’, 2 = ’Thursday’, and 3 = ’Friday’.
If you specify ORDER1=INTERNAL, the rows of the comparative histogram will appear in
day-of-the-week order (Wednesday, Thursday, Friday) from top to bottom.
If you specify ORDER1=FORMATTED, the rows (columns) are displayed from top to bottom
(left to right) in increasing order of the formatted values of the first CLASS= variable. In
the preceding illustration, if you specify ORDER1=FORMATTED, the rows will appear in
alphabetical order (Friday, Thursday, Wednesday) from top to bottom.
If you specify ORDER1=DATA, the rows (columns) are displayed from top to bottom (left to
right) in the order in which the values of the first CLASS= variable first appear in the input data
set.
If you specify ORDER1=FREQ, the rows (columns) are displayed from top to bottom (left to
right) in order of decreasing frequency count. If two or more classes have the same frequency
count, the order is determined by the formatted values.
By default, ORDER1=INTERNAL.
ORDER2=INTERNAL | FORMATTED | DATA | FREQ

specifies the display order for the values of the second CLASS= variable.
The levels of the second CLASS= variable are always constructed using the formatted values of the
variable, and the formatted values are always used to label the columns of a two-way comparative
histogram. You can use the ORDER2= option to determine the order of the columns.
The layout of a two-way comparative histogram is determined by using the ORDER1= option to obtain
the order of the rows from top to bottom (recall that ORDER1=INTERNAL by default). Then the
ORDER2= option is applied to the observations corresponding to the first row to obtain the order of the
columns from left to right. If any columns remain unordered (that is, the categories are unbalanced),
the ORDER2= option is applied to the observations in the second row, and so on, until all the columns
have been ordered.
The values of the ORDER2= option are interpreted as described for the ORDER1= option. By default,
ORDER2=INTERNAL.
OUTHISTOGRAM=SAS-data-set
creates a SAS data set that saves the midpoints or endpoints of the histogram intervals, the observed
percent of observations in each interval, and (optionally) the percent of observations in each interval
estimated from a fitted normal distribution. By default, interval midpoint values are saved in the
variable _MIDPT_. If the ENDPOINTS= option is specified, intervals are identified by endpoint
values instead. If RTINCLUDE is specified, the _MAXPT_ variable contains upper endpoint values.
Otherwise, lower endpoint values are saved in the _MINPT_ variable.
RTINCLUDE
includes the right endpoint of each histogram interval in that interval. The left endpoint is included by
default.
SIGMA=value
specifies the parameter for normal density curves requested with the NORMAL option. Enclose the
SIGMA= option in parentheses after the NORMAL option. The default value is the sample standard
deviation of the observations in the cell.
UPPER=value
specifies the upper bound for a kernel density estimate curve. Enclose the UPPER= option in paren-
theses after the KERNEL option. You can specify a single upper bound or a list of upper bounds. By
default, a kernel density estimate curve has no upper bound.
VSCALE=PERCENT | COUNT | PROPORTION

specifies the scale of the vertical axis. The value COUNT scales the data in units of the number of
observations per data unit. The value PERCENT scales the data in units of percent of observations per
data unit. The value PROPORTION scales the data in units of proportion of observations per data unit.
The default is PERCENT.

BARWIDTH=value
specifies the width of the histogram bars in screen percent units.
CBARLINE=color
specifies the color of the outline of the histogram bars. This option overrides the C= option in the
SYMBOL1 statement.
CFILL=color
specifies a color used to fill the bars of the histograms (or the areas under a fitted curve if you also
specify the FILL option). See the entry for the FILL option for additional details. See Output 6.6.1 and
Example 6.7 for examples. Refer to SAS/GRAPH: Reference for a list of colors. By default, bars and
curve areas are not filled.
CFRAMENLEG=color | EMPTY
CFRAMENLEG
specifies that the legend requested with the NLEGEND option (or the variable _TILELB_ in a
CLASSSPEC= data set) is to be framed and that the frame is to be filled with the color indicated. If
you specify CFRAMENLEG=EMPTY, a frame is drawn but not filled with a color.
CGRID=color
specifies the color for grid lines requested with the GRID option. By default, grid lines are the same
color as the axes. If you use CGRID=, you do not need to specify the GRID option.
CLIPSPEC=CLIP | NOFILL
specifies that histogram bars are clipped at the upper and lower specification limit lines when there are
no observations outside the specification limits. The bar intersecting the lower specification limit is
clipped if there are no observations less than the lower limit; the bar intersecting the upper specification
limit is clipped if there are no observations greater than the upper limit. If you specify CLIPSPEC=CLIP,
the histogram bar is truncated at the specification limit. If you specify CLIPSPEC=NOFILL, the portion
of a filled histogram bar outside the specification limit is left unfilled. Specifying CLIPSPEC=NOFILL
when histogram bars are not filled has no effect.
FRONTREF
draws reference lines requested with the HREF= and VREF= options in front of the histogram bars.
By default, reference lines are drawn behind the histogram bars and can be obscured by them.
HOFFSET=value
specifies the offset in percent screen units at both ends of the horizontal axis. Specify HOFFSET=0 to
eliminate the default offset.
LGRID=n
specifies the line type for the grid requested with the GRID option. If you use the LGRID= option, you
do not need to specify the GRID option. The default is 1, which produces a solid line.
NLEGEND< =’label’ >

specifies the form of a legend that is displayed inside each tile and indicates the sample size of the cell.
The following two forms are available:
If you specify the NLEGEND option, the form is N D n where n is the cell sample size.
If you specify the NLEGEND=’label’ option, the form is label = n where n is the cell sample
size. The label can be up to 16 characters and must be enclosed in quotes. For instance, you
might specify NLEGEND=’Number of Parts’ to request a label of the form Number of Parts = n.
See Figure 6.6 for an example. You can use the CFRAMENLEG= option to frame the sample size
legend. The variable _TILELB_ in a CLASSSPEC= data set overrides the NLEGEND option. By
default, no legend is displayed.
NLEGENDPOS=NW | NE
specifies the position of the legend requested with the NLEGEND option or the variable _TILELB_ in
a CLASSSPEC= data set. If NLEGENDPOS=NW, the legend is displayed in the northwest corner
of the tile; if NLEGENDPOS=NE, the legend is displayed in the northeast corner of the tile. See
Figure 6.6 for an illustration. The default is NE.
PFILL=pattern
specifies a pattern used to fill the bars of the histograms (or the areas under a fitted curve if you also
specify the FILL option). See the entries for the CFILL= and FILL options for additional details. Refer
to SAS/GRAPH: Reference for a list of pattern values. By default, the bars and curve areas are not
filled.
Details: COMPHISTOGRAM Statement F 305
TILELEGLABEL=’label’
specifies a label displayed to the left of the legend that is created when you provide _CTILE_ and
_TILELG_ variables in a CLASSSPEC= data set. The label can be up to 16 characters and must be
enclosed in quotes. The default label is Tiles:.
VOFFSET=value
specifies the offset in percent screen units at the upper end of the vertical axis.
WBARLINE=n
specifies the width of bar outlines. By default, n = 1.
WGRID=n
specifies the width of the grid lines requested with the GRID option. By default, grid lines are the
same width as the axes. If you use the WGRID= option, you do not need to specify the GRID option.
Details: COMPHISTOGRAM Statement

ODS Graphics
destination where the graph is produced. COMPHISTOGRAM options used to control the appearance of
traditional graphics are ignored for ODS Graphics output.
When ODS Graphics is in effect, the COMPHISTOGRAM statement assigns a name to the graph it creates.
You can use this name to reference the graph when using ODS. The name is listed in Table 6.22.
Table 6.22 ODS Graphics Produced by the COMPHISTOGRAM Statement

Histogram Comparative histogram
producing charts.
Examples: COMPHISTOGRAM Statement

This section provides advanced examples of comparative histograms.
Example 6.6: Adding Insets with Descriptive Statistics

N OTE : See Machine Study with Comparative Histogram in the SAS/QC Sample Library.
Three similar machines are used to attach a part to an assembly. One hundred assemblies are sampled from
the output of each machine, and a part position is measured in millimeters. The following statements save the
measurements in a SAS data set named Machines:
data Machines;
input position @@;
label position='Position in Millimeters';
if (_n_ <= 100) then Machine = 'Machine 1';
else if (_n_ <= 200) then Machine = 'Machine 2';
else Machine = 'Machine 3';
datalines;
-0.17 -0.19 -0.24 -0.24 -0.12 0.07 -0.61 0.22 1.91 -0.08
-0.59 0.05 -0.38 0.82 -0.14 0.32 0.12 -0.02 0.26 0.19
-0.07 0.13 -0.49 0.07 0.65 0.94 -0.51 -0.61 -0.57 -0.51
0.01 -0.51 0.07 -0.16 -0.32 -0.42 -0.42 -0.34 -0.34 -0.35
-0.49 0.11 -0.42 0.76 0.02 -0.59 -0.28 1.12 -0.02 -0.60
-0.64 0.13 -0.32 -0.77 -0.02 -0.07 -0.49 -0.53 -0.22 0.61
-0.23 0.02 0.53 0.23 -0.44 -0.05 0.37 -0.42 0.70 -0.35
... more lines ...
0.58 0.46 0.58 0.92 0.70 0.81 0.07 0.33 0.82 0.62
0.48 0.41 0.78 0.58 0.43 0.07 0.27 0.49 0.79 0.92
0.79 0.66 0.22 0.71 0.53 0.57 0.90 0.48 1.17 1.03
;
Distinct specification limits for the three machines are provided in a data set named speclims.
data speclims;
input Machine $9. _lsl_ _usl_;
_var_ = 'position';
datalines;
Machine 1 -0.5 0.5
Machine 2 0.0 1.0
Machine 3 0.0 1.0
;
The following statements create a comparative histogram for the measurements in Machines that displays the
specification limits in speclims.
proc capability data=Machines noprint;

spec cleft cright;
comphist position / class = Machine
nrows = 3
intertile = 1
Examples: COMPHISTOGRAM Statement F 307
midpoints = -1.2 to 2.2 by 0.1

kernel(fill)
classspecs = speclims;
inset mean std="Std Dev" / pos = ne format = 6.3;
run;
The display is shown in Output 6.6.1.
Output 6.6.1 Comparative Histogram
The INSET statement is used to inset the sample mean and standard deviation for each machine in the
corresponding tile. The MIDPOINTS= option specifies the midpoints of the histogram bins. Kernel density
estimates are displayed using the KERNEL option. The curve areas outside the specification limits are filled
using the CLEFT and CRIGHT options in the SPEC statement, and the area between the limits is filled using
the CFILL= option in COMPHISTOGRAM statement.
Example 6.7: Creating a Two-Way Comparative Histogram

N OTE : See Two-Way Comparative Histogram in the SAS/QC Sample Library.
Two suppliers (A and B) provide disk drives for a computer manufacturer. The manufacturer measures the
disk drive opening width to compare the process capabilities of the suppliers and determine whether there
has been an improvement from 1992 to 1993.
The following statements save the measurements in a data set named Disk. There are two classification
variables, Supplier and Year, and a format is associated with Year.
proc format ;
value mytime 1 = '1992'
2 = '1993' ;
data disk;
input @1 supplier $10. year width;
label width = 'Opening Width (inches)';
format year mytime.;
datalines;
Supplier A 1 1.8932
Supplier A 1 1.8952
. . .
. . .
Supplier B 1 1.8980
Supplier B 1 1.8986
Supplier A 2 1.8978
Supplier A 2 1.8966
. . .
. . .
Supplier B 2 1.8967
Supplier B 2 1.8997
;
The following statements create the comparative histogram in Output 6.7.1:
* Define a format for time periods;

proc format ;
value mytime
1 = '1992'
2 = '1993'
3 = 'Target for 1994'
;
* Simulate the data;
data Disk;
keep Supplier Year Width;
length Supplier $ 16;
label Width = 'Opening Width (inches)';
format Year mytime. ;
Year = 1;
Supplier = 'Supplier A';
avg = 1.895 ;
std = 0.0027 ;
do i = 1 to 260;
Width = avg + std * rannor(15535); output;
Examples: COMPHISTOGRAM Statement F 309
end;
Supplier = 'Supplier B';
avg = 1.8983 ;
std = 0.0024 ;
do i = 1 to 260;
end;
Year = 2;
Supplier = 'Supplier A';
avg = 1.8970 ;
std = 0.0013 ;
do i = 1 to 260;
end;
Supplier = 'Supplier B';
avg = 1.8980 ;
std = 0.0013 ;
do i = 1 to 260;
end;
run;
title "Results of Supplier Training Program";

proc capability data=Disk noprint;
specs lsl = 1.8925
usl = 1.9027;
comphist Width / class = ( Supplier Year )
classkey = ('Supplier A' '1993')
intertile = 1.0
vaxis = 0 10 20 30
ncols = 2
nrows = 2
odstitle = title;
inset cpk (4.2) / noframe pos = n;
run;
Output 6.7.1 Two-Way Comparative Histogram
The CLASSKEY= option specifies the key cell as the observations for which Supplier is equal to ‘SUPPLIER
A’ and Year is equal to 2. This cell determines the binning for the other cells, and (because the NOKEYMOVE
option is not specified) the columns are interchanged so that this cell is displayed in the upper left corner.
Note that if the CLASSKEY= option were not specified, the default key cell would be the observations
for which Supplier is equal to ‘SUPPLIER A’ and Year is equal to 1. If the CLASSKEY= option were not
specified (or if the NOKEYMOVE option were specified), the column labeled 1992 would be displayed to
the left of the column labeled 1993. See the entry for the CLASSKEY= option on page 296 for details.
The VAXIS= option specifies the tick mark labels for the vertical axis, while NROWS=2 and NCOLS=2
specify a 2 2 arrangement for the tiles. The INSET statement is used to display the capability index Cpk
for each cell. Output 6.7.1 provides evidence that both suppliers have reduced variability from 1992 to 1993.
HISTOGRAM Statement: CAPABILITY Procedure F 311
HISTOGRAM Statement: CAPABILITY Procedure
Overview: HISTOGRAM Statement

Histograms are typically used in process capability analysis to compare the distribution of measurements
from an in-control process with its specification limits. In addition to creating histograms, you can use the
HISTOGRAM statement to do the following:
specify the midpoints or endpoints for histogram intervals
display specification limits on histograms
display density curves for fitted theoretical distributions on histograms
request goodness-of-fit tests for fitted distributions
display kernel density estimates on histograms
inset summary statistics and process capability indices on histograms
save histogram intervals and parameters of fitted distributions in output data sets
create hanging histograms
create comparative histograms by using the HISTOGRAM statement together with a CLASS statement
You have three alternatives for producing histograms with the HISTOGRAM statement:
statement.
Getting Started: HISTOGRAM Statement

This section introduces the HISTOGRAM statement with examples that illustrate commonly used options.
Complete syntax for the HISTOGRAM statement is presented in the section “Syntax: HISTOGRAM
Statement” on page 318, and advanced examples are given in the section “Examples: HISTOGRAM
Statement” on page 376.
Creating a Histogram with Specification Limits

N OTE : See Histogram with Fitted Normal Curve in the SAS/QC Sample Library.
A semiconductor manufacturer produces printed circuit boards that are sampled to determine whether
the thickness of their copper plating lies between a lower specification limit of 3.45 mils and an upper
specification limit of 3.55 mils. The plating process is assumed to be in statistical control. The plating
thicknesses of 100 boards are saved in a data set named Trans, created by the following statements:
data Trans;
input Thick @@;
label Thick='Plating Thickness (mils)';
datalines;
3.468 3.428 3.509 3.516 3.461 3.492 3.478 3.556 3.482 3.512
3.490 3.467 3.498 3.519 3.504 3.469 3.497 3.495 3.518 3.523
3.458 3.478 3.443 3.500 3.449 3.525 3.461 3.489 3.514 3.470
3.561 3.506 3.444 3.479 3.524 3.531 3.501 3.495 3.443 3.458
3.481 3.497 3.461 3.513 3.528 3.496 3.533 3.450 3.516 3.476
3.512 3.550 3.441 3.541 3.569 3.531 3.468 3.564 3.522 3.520
3.505 3.523 3.475 3.470 3.457 3.536 3.528 3.477 3.536 3.491
3.510 3.461 3.431 3.502 3.491 3.506 3.439 3.513 3.496 3.539
3.469 3.481 3.515 3.535 3.460 3.575 3.488 3.515 3.484 3.482
3.517 3.483 3.467 3.467 3.502 3.471 3.516 3.474 3.500 3.466
;
The following statements create the histogram shown in Figure 6.8:
title 'Process Capability Analysis of Plating Thickness';

proc capability data=Trans noprint;
spec lsl = 3.45 usl = 3.55;
histogram Thick / odstitle = title;
run;
A histogram is created for each variable listed after the keyword HISTOGRAM. The SPEC statement, which
is optional, provides the specification limits that are displayed on the histogram. For more information about
the SPEC statement, see “SPEC Statement” on page 225.
The NOPRINT option suppresses printed output with summary statistics for the variable Thick that would be
displayed by default. See “Computing Descriptive Statistics” on page 207 for an example of this output.
Getting Started: HISTOGRAM Statement F 313
Figure 6.8 Histogram Created with Traditional Graphics
Adding a Normal Curve to the Histogram

This example is a continuation of the preceding example.
The following statements fit a normal distribution from the thickness measurements and superimpose the
fitted density curve on the histogram:
proc capability data=Trans;

spec lsl = 3.45 usl = 3.55;
histogram / normal;
run;
The ODS GRAPHICS ON statement specified before the PROC CAPABILITY statement enables ODS
Graphics, so the histogram is created using ODS Graphics instead of traditional graphics.
The NORMAL option summarizes the fitted distribution in the printed output shown in Figure 6.9, and it
specifies that the normal curve be displayed on the histogram shown in Figure 6.10.
Figure 6.9 Summary for Fitted Normal Distribution

Process Capability Analysis of Plating Thickness

Fitted Normal Distribution for Thick (Plating Thickness (mils))
Parameters for Normal

Distribution
Parameter Symbol Estimate
Mean Mu 3.49533
Std Dev Sigma 0.032117
Goodness-of-Fit Tests for Normal Distribution

Test Statistic DF p Value
Chi-Square Chi-Sq 6.96953022 5 Pr > Chi-Sq 0.223
Percent Outside Specifications for Normal

Distribution
Lower Limit Upper Limit
LSL 3.450000 USL 3.550000
Obs Pct < LSL 8.000000 Obs Pct > USL 5.000000
Est Pct < LSL 7.906248 Est Pct > USL 4.435722
Quantiles for Normal

Distribution
Quantile
Percent Observed Estimated
1.0 3.42950 3.42061
5.0 3.44300 3.44250
10.0 3.45750 3.45417
25.0 3.46950 3.47367
50.0 3.49600 3.49533
75.0 3.51650 3.51699
90.0 3.53550 3.53649
95.0 3.55300 3.54816
99.0 3.57200 3.57005
Figure 6.10 Histogram Superimposed with Normal Curve
The printed output includes the following:
parameters for the normal curve. The normal parameters and are estimated by the sample mean
(O D 3:49533) and the sample standard deviation (O D 0:032117).
goodness-of-fit tests based on the empirical distribution function (EDF): the Anderson-Darling, Cramér–
von Mises, and Kolmogorov-Smirnov tests. The p-values for these tests are greater than the usual
cutoff values of 0.05 and 0.10, indicating that the thicknesses are normally distributed.
a chi-square goodness-of-fit test. The p-value of 0.223 for this test indicates that the thicknesses are
normally distributed. In general EDF tests (when available) are preferable to chi-square tests. See the
section “EDF Goodness-of-Fit Tests” on page 364 for details.
observed and estimated percentages outside the specification limits
observed and estimated quantiles
For details, including formulas for the goodness-of-fit tests, see “Printed Output” on page 362. Note that the
NOPRINT option in the PROC CAPABILITY statement suppresses only the printed output with summary
statistics for the variable Thick. To suppress the printed output in Figure 6.9, specify the NOPRINT option
enclosed in parentheses after the NORMAL option as in “Customizing a Histogram” on page 316.
The NORMAL option is one of many options that you can specify in the HISTOGRAM statement. See
the section “Syntax: HISTOGRAM Statement” on page 318 for a complete list of options or the section
“Dictionary of Options” on page 326 for detailed descriptions of options.
Customizing a Histogram
This example is a continuation of the preceding example. The following statements show how you can use
HISTOGRAM statement options and INSET statements to customize a histogram:
title 'Process Capability Analysis of Plating Thickness';

proc capability data=Trans noprint;
spec lsl = 3.45 usl = 3.55;
histogram Thick / normal
midpoints = 3.4 to 3.6 by 0.025
vscale = count
odstitle = title
nospeclegend;
inset lsl usl;
inset n mean (5.2) cpk (5.2);
run;
The histogram is displayed in Figure 6.11.
Figure 6.11 Customizing the Appearance of the Histogram
The MIDPOINTS= option specifies a list of values to use as bin midpoints. The VSCALE=COUNT option
requests a vertical axis scaled in counts rather than percents. The INSET statements inset the specification
limits and summary statistics. The NOSPECLEGEND option suppress the default legend for the specification
limits that is shown in Figure 6.8.
For more information about HISTOGRAM statement options, see the section “Dictionary of Options” on
page 326. For details on the INSET statement, see “INSET Statement: CAPABILITY Procedure” on
page 398.
Syntax: HISTOGRAM Statement

The syntax for the HISTOGRAM statement is as follows:
HISTOGRAM < variables > < / options > ;
You can specify the keyword HIST as an alias for HISTOGRAM. You can use any number of HISTOGRAM
statements after a PROC CAPABILITY statement. The components of the HISTOGRAM statement are
described as follows.
variables
are the process variables for which histograms are to be created. If you specify a VAR statement,
the variables must also be listed in the VAR statement. Otherwise, the variables can be any numeric
variables in the input data set. If you do not specify variables in a VAR statement or in the HISTOGRAM
statement, then by default, a histogram is created for each numeric variable in the DATA= data set. If
you use a VAR statement and do not specify any variables in the HISTOGRAM statement, then by
default, a histogram is created for each variable listed in the VAR statement.
For example, suppose a data set named steel contains exactly two numeric variables named length and
width. The following statements create two histograms, one for length and one for width:

histogram;
run;
The following statements also create histograms for length and width:

var length width;
histogram;
run;
The following statements create a histogram for length only:

var length width;
histogram length;
run;
options
add features to the histogram. Specify all options after the slash (/) in the HISTOGRAM statement.
For example, in the following statements, the NORMAL option displays a fitted normal curve on the
histogram, the MIDPOINTS= option specifies midpoints for the histogram, and the CTEXT= option
specifies the color of the text:
Syntax: HISTOGRAM Statement F 319

histogram length / normal
midpoints = 5.6 5.8 6.0 6.2 6.4
ctext = yellow;
run;
Summary of Options
The following tables list the HISTOGRAM statement options by function. For detailed descriptions, see
Parametric Density Estimation Options

Table 6.23 lists options that display a parametric density estimate on the histogram.
Table 6.23 Parametric Distribution Options
Option Description
BETA(beta-options) Fits beta distribution with threshold
parameter , scale parameter , and
shape parameters ˛ and ˇ
EXPONENTIAL(exponential-options) Fits exponential distribution with
threshold parameter and scale
parameter
GAMMA(gamma-options) Fits gamma distribution with threshold
shape parameter ˛
GUMBEL(Gumbel-options) Plots Gumbel distribution with location
IGAUSS(iGauss-options) Plots inverse Gaussian distribution with
mean and shape parameter
LOGNORMAL(lognormal-options) Fits lognormal distribution with threshold
parameter
NORMAL(normal-options) Fits normal distribution with mean and
standard deviation
PARETO(Pareto-options) Plots Pareto distribution with threshold
shape parameter ˛
POWER(power-options) Plots power function distribution with
RAYLEIGH(Rayleigh-options) Plots Rayleigh distribution with threshold
SB(SB-options) Fits Johnson SB distribution with
and shape parameters ı and
Option Description
SU(SU-options) Fits Johnson SU distribution with
location parameter , scale parameter ,
and shape parameters ı and
WEIBULL(Weibull-options) Fits Weibull distribution with threshold
shape parameter c
Table 6.24 lists secondary options that specify parameters for fitted parametric distributions and that control
the display of fitted curves. Specify these secondary options in parentheses after the distribution keyword.
For example, the following statements fit a normal curve by using the NORMAL option:
proc capability;
histogram / normal(color=red mu=10 sigma=0.5);
run;
The COLOR= normal-option draws the curve in red, and the MU= and SIGMA= normal-options specify the
parameters D 10 and D 0:5 for the curve. Note that the sample mean and sample standard deviation are
used to estimate and , respectively, when the MU= and SIGMA= options are not specified.
You can specify lists of values for distribution parameters to display more than one fitted curve from the same
distribution family on a histogram. Option values are matched by list position. You can specify the value
EST in a list of distribution parameter values to use an estimate of the parameter.
For example, the following code displays two normal curves on a histogram:
proc capability;
histogram / normal(color=(red blue) mu=10 est sigma=0.5 est);
run;
The first curve is red, with D 10 and D 0:5. The second curve is blue, with equal to the sample mean
and equal to the sample standard deviation.
See the section “Formulas for Fitted Curves” on page 349 for detailed information about the families of
parametric distributions that you can fit with the HISTOGRAM statement.
Option Description
Options Used with All Parametric Distributions
COLOR= Specifies color of fitted density curve
FILL Fills area under fitted density curve
INDICES Calculates capability indices based on fitted distribution
L= Specifies line type of fitted curve
MIDPERCENTS Prints table of midpoints of histogram intervals
NOPRINT Suppresses printed output summarizing fitted curve
PERCENTS= Lists percents for which quantiles calculated from data and
quantiles estimated from fitted curve are tabulated
Option Description
SYMBOL= Specifies character used for fitted density curve in line printer plots
W= Specifies width of fitted density curve
Beta-Options
ALPHA= Specifies first shape parameter ˛ for fitted beta curve
BETA= Specifies second shape parameter ˇ for fitted beta curve
SIGMA= Specifies scale parameter for fitted beta curve
THETA= Specifies lower threshold parameter for fitted beta curve
Exponential-Options
SIGMA= Specifies scale parameter for fitted exponential curve
THETA= Specifies threshold parameter for fitted exponential curve
Gamma-Options
ALPHA= Specifies shape parameter ˛ for fitted gamma curve
ALPHADELTA= Specifies change in successive estimates of ˛ at which the
Newton-Raphson approximation of ˛O terminates
ALPHAINITIAL= Specifies initial value for ˛ in Newton-Raphson approximation of
˛O
MAXITER= Specifies maximum number of iterations in Newton-Raphson
approximation of ˛O
SIGMA= Specifies scale parameter for fitted gamma curve
THETA= Specifies threshold parameter for fitted gamma curve
Gumbel-Options
EDFNSAMPLES= Specifies number of samples for EDF goodness-of-fit simulation
EDFSEED= Specifies seed value for EDF goodness-of-fit simulation
MU= Specifies location parameter for fitted Gumbel curve
SIGMA= Specifies scale parameter for fitted Gumbel curve
IGauss-Options
LAMBDA= Specifies shape parameter for fitted inverse Gaussian curve
MU= Specifies mean for fitted inverse Gaussian curve
Lognormal-Options
SIGMA= Specifies shape parameter for fitted lognormal curve
THETA= Specifies threshold parameter for fitted lognormal curve
ZETA= Specifies scale parameter for fitted lognormal curve
Normal-Options
MU= Specifies mean for fitted normal curve
SIGMA= Specifies standard deviation for fitted normal curve
Option Description
Pareto-Options
ALPHA= Specifies shape parameter ˛ for fitted Pareto curve
SIGMA= Specifies scale parameter for fitted Pareto curve
THETA= Specifies threshold parameter for fitted Pareto curve
Power-Options
ALPHA= Specifies shape parameter ˛ for fitted power function curve
SIGMA= Specifies scale parameter for fitted power function curve
THETA= Specifies threshold parameter for fitted power function curve
Rayleigh-Options
SIGMA= Specifies scale parameter for fitted Rayleigh curve
THETA= Specifies threshold parameter for fitted Rayleigh curve
SB -Options
DELTA= Specifies first shape parameter ı for fitted SB curve
FITINTERVAL= Specifies z-value for method of percentiles
FITMETHOD= Specifies method of parameter estimation
FITTOLERANCE= Specifies tolerance for method of percentiles
GAMMA= Specifies second shape parameter for fitted SB curve
SIGMA= Specifies scale parameter for fitted SB curve
THETA= Specifies lower threshold parameter for fitted SB curve
SU -Options
DELTA= Specifies first shape parameter ı for fitted SU curve
FITINTERVAL= Specifies z-value for method of percentiles
FITMETHOD= Specifies method of parameter estimation
FITTOLERANCE= Specifies tolerance for method of percentiles
GAMMA= Specifies second shape parameter for fitted SU curve
OPTBOUNDRANGE= Specifies the sampling range for parameter starting values in MLE
optimization
OPTMAXITER= Specifies an iteration limit for MLE optimization
OPTMAXSTARTS= Specifies the maximum number of starting points to be used for
MLE optimization
OPTPRINT Prints an iteration history for MLE optimization
OPTSEED= Specifies a seed value for MLE optimization
OPTTOLERANCE= Specifies the optimality tolerance for MLE optimization
SIGMA= Specifies scale parameter for fitted SU curve
THETA= Specifies location parameter for fitted SU curve
Option Description
Weibull-Options
C= Specifies shape parameter c for fitted Weibull curve
CDELTA= Specifies change in successive estimates of c at which the
Newton-Raphson approximation of cO terminates
CINITIAL= Specifies initial value for c in Newton-Raphson approximation of cO
MAXITER= Specifies maximum number of iterations in Newton-Raphson
approximation of cO
SIGMA= Specifies scale parameter for fitted Weibull curve
THETA= Specifies threshold parameter for fitted Weibull curve
Nonparametric Density Estimation Options
Table 6.25 Kernel Density Estimation Options
Option Description
KERNEL(kernel-options) Fits kernel density estimates
Specify the options listed in Table 6.26 in parentheses after the keyword KERNEL to control features of
kernel density estimates requested with the KERNEL option.
Table 6.26 Kernel-Options
Option Description
C= Specifies standardized bandwidth parameter c for fitted kernel
density estimate
COLOR= Specifies color of the fitted kernel density curve
FILL Fills area under fitted kernel density curve
K= Specifies type of kernel function
L= Specifies line type used for fitted kernel density curve
LOWER= Specifies lower bound for fitted kernel density curve
SYMBOL= Specifies character used for fitted kernel density curve in line
printer plots
UPPER= Specifies upper bound for fitted kernel density curve
W= Specifies line width for fitted kernel density curve
General Options
Table 6.27 summarizes general options for the HISTOGRAM statement, including options for enhancing
charts and producing output data sets.
Table 6.27 General HISTOGRAM Statement Options
Option Description
Options to Create Output Data Sets
OUTFIT= Requests information about fitted curves
OUTHISTOGRAM= Requests information about histogram intervals
OUTKERNEL= Creates a data set containing kernel density estimates
General Histogram Layout Options

CLIPCURVES Scales vertical axis without considering fitted curves
CONTENTS= Specifies table of contents entry for histogram grouping
CURVELEGEND= Specifies LEGEND statement for curves
ENDPOINTS= Lists endpoints for histogram intervals
HANGING Constructs hanging histogram
MIDPERCENTS Prints table of histogram intervals
MIDPOINTS= Lists midpoints for histogram intervals
NENDPOINTS= Specifies number of histogram interval endpoints
NMIDPOINTS= Specifies number of histogram interval midpoints
NOBARS Suppresses histogram bars
NOCURVELEGEND Suppresses legend for curves
NOPLOT Suppresses plot
NOSPECLEGEND Suppresses specifications legend
NOTABCONTENTS Suppresses table of contents entries for tables produced by
HISTOGRAM statement
RTINCLUDE Includes right endpoint in interval
SPECLEGEND= Specifies LEGEND statement for specification limits
Options to Enhance Graphical Output

BARLABEL= Produces labels above histogram bars
BARWIDTH= Specifies width for the bars
BMCFILL= Specifies fill color for box-and-whisker plot in bottom margin
BMCFRAME= Specifies fill color bottom margin plot frame
BMCOLOR= Specifies color for bottom margin plot
BMMARGIN= Specifies height of margin for bottom margin plot
BMPLOT= Requests a plot in bottom margin of histogram
CBARLINE= Specifies color for outlines of histogram bars
CFILL= Specifies color for filling under curve
Option Description
CLIPREF Draws reference lines behind histogram bars
CLIPSPEC= Clips histogram bars at specification limits
DESCRIPTION= Specifies description for plot in graphics catalog
FRONTREF Draws reference lines in front of histogram bars
GRID Creates a grid
HMINOR= Specifies number of horizontal minor tick marks
HOFFSET= Specifies offset for horizontal axis
INTERBAR= Specifies space between histogram bars
LEGEND= Identifies LEGEND statement
LGRID= Specifies a line type for grid lines
MAXNBIN= Specifies maximum number of bins to display
MAXSIGMAS= Limits the number of bins that display to within a specified
number of standard deviations above and below mean of data
in key cell
MIDPOINTS= Specifies midpoints for histogram intervals
PFILL= Specifies pattern for filling under curve
STATREFSUBCHAR= Specifies substitution character for displaying statistic values
in STATREFLABELS= labels
TURNVLABELS Turns and vertically strings out characters in labels for vertical
axis
VAXIS= Specifies AXIS statement or values for vertical axis
VMINOR= Specifies number of vertical minor tick marks
VOFFSET= Specifies length of offset at upper end of vertical axis
Option Description
WBARLINE= Specifies line thickness for bar outlines

BARFILL= Enables bars in different cells of a comparative histogram to be
filled with different colors
ODSFOOTNOTE= Specifies footnote displayed on histogram
ODSFOOTNOTE2= Specifies secondary footnote displayed on histogram
ODSTITLE= Specifies title displayed on histogram
ODSTITLE2= Specifies secondary title displayed on histogram

ANNOKEY Applies annotation requested in ANNOTATE= data set to key
cell only
CTEXTSIDE= Specifies color for row labels of comparative histograms
CTEXTTOP= Specifies color for column labels of comparative histograms
NCOLS= Specifies number of columns in comparative histogram
NROWS= Specifies number of rows in comparative histogram
Options to Enhance Line Printer Plots

The following sections provide detailed descriptions of options specific to the HISTOGRAM statement. See
General Options
ALPHA=value-list
specifies the shape parameter ˛ for fitted curves requested with the BETA, GAMMA, PARETO, and
POWER options. Enclose the ALPHA= option in parentheses after the distribution keyword. If you do
not specify a value for ˛, the procedure calculates a maximum likelihood estimate. See Example 6.8.
You can specify A= as an alias for ALPHA= if you use it as a beta-option. You can specify SHAPE=
as an alias for ALPHA= if you use it as a gamma-option.
BARFILL=variable-list
specifies one or more variables whose values determine the colors of the bars in the cells of a
comparative histogram. Cells that are associated with a particular value of a variable in the variable-list
are the same color. The colors that are used are determined by the ODS style. If the HISTOGRAM
statement applies to more than one analysis variable, you can specify more than one variable in the
variable-list , and those variables are matched with analysis variables by their positions in the list.
N OTE : This option applies only when ODS Graphics is enabled.
BARLABEL=COUNT | PERCENT | PROPORTION

displays labels above the histogram bars. If you specify BARLABEL=COUNT, the label shows the
number of observations associated with a particular bar. BARLABEL=PERCENT shows the percent
of observations represented by that bar. If you specify BARLABEL=PROPORTION, the label displays
the proportion of observations associated with the bar.
BETA< (beta-options ) >

displays a fitted beta density curve on the histogram. The curve equation is
(
.x /˛ 1 . C x/ˇ 1
B.˛;ˇ / .˛Cˇ 1/
hv for < x < C
p.x/ D
0 for x or x C
.˛/.ˇ /
where B.˛; ˇ/ D .˛Cˇ /
and
D lower threshold parameter (lower endpoint parameter)

and
8
:
The beta distribution is bounded below by the parameter and above by the value C . You can
specify and by using the THETA= and SIGMA= beta-options. The following statements fit a beta
distribution bounded between 50 and 75 by using maximum likelihood estimates for ˛ and ˇ:
proc capability;
histogram length / beta(theta=50 sigma=25);
run;
In general, the default values for THETA= and SIGMA= are 0 and 1, respectively. You can specify
THETA=EST and SIGMA=EST to request maximum likelihood estimates for and .
The beta distribution has two shape parameters, ˛ and ˇ. If these parameters are known, you can
specify their values with the ALPHA= and BETA= beta-options. If you do not specify values, the
procedure calculates maximum likelihood estimates for ˛ and ˇ.
The BETA option can appear only once in a HISTOGRAM statement. Table 6.24 lists secondary
options you can specify with the BETA option. See Example 6.8. Also see “Formulas for Fitted Curves”
on page 349.
BETA=value-list
B=value-list
specifies the second shape parameter ˇ for beta density curves requested with the BETA option.
Enclose the BETA= option in parentheses after the BETA option. If you do not specify a value for ˇ,
the procedure calculates a maximum likelihood estimate. See Example 6.8.
BMPLOT=CARPET | DOTPLOT | SKELETAL | SCHEMATIC

produces a carpet plot, dot plot, or box-and-whisker plot along the bottom margin of a histogram. A
carpet plot or dot plot shows the distribution of individual observations along the histogram’s horizontal
axis. A carpet plot represents each observation with a vertical line. A dot plot marks each observation
with a symbol. A box-and-whisker plot gives a summary of the data distribution that a histogram alone
does not provide. The left and right edges of the box are located at the first and third quartiles. A
central vertical line is drawn at the median and a symbol is plotted inside the box at the mean. If you
specify the SKELETAL keyword, a box-and-whisker plot is produced with whiskers extending to the
minimum and maximum values. If you specify SCHEMATIC, a schematic box-and-whisker plot is
produced. In a schematic box-and-whisker plot, the whiskers extend to the smallest value within the
lower fence and the largest value within the upper fence. Fences are defined in terms of the interquartile
range (IQR). The lower fence is 1.5 IQR below the first quartile and the upper fence is 1.5 IQR above
the third quartile. Each observation outside the fences is plotted with a symbol.
C=value-list
specifies the shape parameter c for Weibull density curves requested with the WEIBULL option.
Enclose the C= option in parentheses after the WEIBULL option. If you do not specify a value for
c, the procedure calculates a maximum likelihood estimate. See Example 6.9. You can specify the
SHAPE= option as an alias for the C= option.
C=value-list | MISE
specifies the standardized bandwidth parameter c for kernel density estimates requested with the
KERNEL option. Enclose the C= option in parentheses after the KERNEL option. You can specify up
to five values to request multiple estimates. You can also specify the C=MISE option, which produces
the estimate with a bandwidth that minimizes the approximate mean integrated square error (MISE).
For example, the following statements compute three density estimates:
proc capability;
histogram length / kernel(c=0.5 1.0 mise);
run;
The first two estimates have standardized bandwidths of 0.5 and 1.0, respectively, and the third has a
bandwidth that minimizes the approximate MISE.
You can also use the C= option with the K= option, which specifies the kernel function, to compute
multiple estimates. If you specify more kernel functions than bandwidths, the last bandwidth in the
list is repeated for the remaining estimates. Likewise, if you specify more bandwidths than kernel
functions, the last kernel function is repeated for the remaining estimates. For example, the following
statements compute three density estimates:
proc capability;
histogram length / kernel(c=1 2 3 k=normal quadratic);
run;
The first uses a normal kernel and a bandwidth of 1, the second uses a quadratic kernel and a bandwidth
of 2, and the third uses a quadratic kernel and a bandwidth of 3. See Example 6.12.
If you do not specify a value for c, the bandwidth that minimizes the approximate MISE is used for all
the estimates.
CLIPCURVES
scales the vertical axis without taking fitted curves into consideration. Curves that extend above the
tallest histogram bar may be clipped. You can use this option to avoid compression of the histogram
bars due to extremely high fitted curve peaks.
DELTA=value-list
specifies the first shape parameter ı for Johnson SB and Johnson SU density curves requested with the
SB and SU options. Enclose the DELTA= option in parentheses after the SB or SU option. If you do
not specify a value for ı, the procedure calculates an estimate.
EDFNSAMPLES=value
specifies the number of simulation samples used to compute p-values for EDF goodness-of-fit statistics
for density curves requested with the GUMBEL, IGAUSS, PARETO, and RAYLEIGH options. Enclose
the EDFNSAMPLES= option in parentheses after the distribution option. The default value is 500.
EDFSEED=value
specifies an integer value used to start the pseudo-random number generator when creating simulation
samples for computing EDF goodness-of-fit statistic p-values for density curves requested with
the GUMBEL, IGAUSS, PARETO, and RAYLEIGH options. Enclose the EDFSEED= option in
parentheses after the distribution option. By default, the procedure uses a random number seed
generated from reading the time of day from the computer’s clock.
ENDPOINTS
ENDPOINTS=value-list
specifies that histogram interval endpoints, rather than midpoints, are aligned with horizontal axis tick
marks. If you specify ENDPOINTS, the number of histogram intervals is based on the number of
observations by using the method of Terrell and Scott (1985). If you specify ENDPOINTS=value-list,
the values must be listed in increasing order and must be evenly spaced. All observations in the input
data set, as well as any specification limits, must lie between the first and last values specified. The
same value-list is used for all variables.
EXPONENTIAL< (exponential-options ) >

EXP< (exponential-options ) >
displays a fitted exponential density curve on the histogram. The curve equation is
hv x
p.x/ D exp. . // for x
0 for x <
where
and
8
:
THETA= exponential-option. The default value for is zero. If you specify THETA=EST, a maximum
likelihood estimate is computed for . You can specify with the SIGMA= exponential-option. By
default, a maximum likelihood estimate is computed for . For example, the following statements fit
an exponential curve with D 10 and with a maximum likelihood estimate for :
proc capability;
histogram / exponential(theta=10 l=2 color=red);
run;
The curve is red and has a line type of 2. The EXPONENTIAL option can appear only once in a
HISTOGRAM statement. Table 6.24 lists secondary options you can specify with the EXPONENTIAL
option. See “Formulas for Fitted Curves” on page 349.
FILL
fills areas under a parametric density curve or kernel density estimate with colors and patterns. Enclose
the FILL option in parentheses after a curve option or the KERNEL option, as in the following
statements:
proc capability;
histogram length / normal(fill) cfill=green pfill=solid;
run;
Depending on the area to be filled (outside or between the specification limits), you can specify the
color and pattern with options in the SPEC statement and HISTOGRAM statement, as summarized in
the following table:
Area Under Curve Statement Color Option Pattern Option

Between specification limits HISTOGRAM CFILL= PFILL=
Left of lower specification limit SPEC CLEFT= PLEFT=
Right of upper specification limit SPEC CRIGHT= PRIGHT=
If you do not display specification limits, the CFILL= and PFILL= options specify the color and pattern
for the entire area under the curve. Solid fills are used by default if patterns are not specified. You
can specify the FILL option with only one fitted curve. For an example, see Output 6.8.1. Refer to
SAS/GRAPH: Reference for a list of available patterns and colors. If you do not specify the FILL option
but specify the options in the preceding table, the colors and patterns are applied to the corresponding
areas under the histogram.
FITINTERVAL=value
specifies the value of z for the method of percentiles when this method is used to fit a Johnson SB or
Johnson SU distribution. The FITINTERVAL= option is specified in parentheses after the SB or SU
option. The default of z is 0.524.
FITMETHOD=PERCENTILE | MLE | MOMENTS

specifies the method used to estimate the parameters of a Johnson SB or Johnson SU distribution. The
FITMETHOD= option is specified in parentheses after the SB or SU option. By default, the method
of percentiles is used. You can specify the MLE keyword to request maximum likelihood estimation.
The OPTBOUNDRANGE=, OPTMAXITER=, OPTMAXSTARTS=, OPTPRINT, OPTSEED=, and
OPTTOLERANCE= options control the optimizer that performs the maximum likelihood calculation.
FITTOLERANCE=value
specifies the tolerance value for the ratio criterion when the method of percentiles is used to fit a
Johnson SB or Johnson SU distribution. The FITTOLERANCE= option is specified in parentheses
after the SB or SU option. The default value is 0.01.

displays a fitted gamma density curve on the histogram. The curve equation is
(
hv
.˛/
. x /˛ 1 exp. . x // for x >
p.x/ D
0 for x
where
and
8
:
The parameter for the gamma distribution must be less than the minimum data value. You can
specify with the THETA= gamma-option. The default value for is 0. If you specify THETA=EST,
a maximum likelihood estimate is computed for . In addition, the gamma distribution has a shape
parameter ˛ and a scale parameter . You can specify these parameters with the ALPHA= and
SIGMA= gamma-options. By default, maximum likelihood estimates are computed for ˛ and .
For example, the following statements fit a gamma curve with D 4 and with maximum likelihood
estimates for ˛ and :
proc capability;
histogram length / gamma(theta=4);
run;
Note that the maximum likelihood estimate of ˛ is calculated iteratively using the Newton-Raphson
approximation. The ALPHADELTA=, ALPHAINITIAL=, and MAXITER= gamma-options control
the approximation.
The GAMMA option can appear only once in a HISTOGRAM statement. Table 6.24 lists secondary
options you can specify with the GAMMA option. See Example 6.9 and “Formulas for Fitted Curves”
on page 349.
GAMMA=value-list
specifies the second shape parameter for Johnson SB and Johnson SU density curves requested with
the SB and SU options. Enclose the GAMMA= option in parentheses after the SB or SU option. If
you do not specify a value for , the procedure calculates an estimate.
GRID
adds a grid to the histogram. Grid lines are horizontal lines positioned at major tick marks on the
vertical axis.

displays a fitted Gumbel (also known as Type 1 extreme value distribution) density curve on the
histogram. The curve equation is
hv .x /=

.x /=

p.x/ D e exp e

where
and
8
:
You can specify values for and with the MU= and SIGMA= Gumbel-options. By default, maximum
likelihood estimates are computed for and .
The GUMBEL option can appear only once in a HISTOGRAM statement. Table 6.24 lists secondary
options you can specify with the GUMBEL option. See “Formulas for Fitted Curves” on page 349.
HANGING
HANG
requests a hanging histogram , as illustrated in Figure 6.12.
Figure 6.12 Hanging Histogram
You can use the HANGING option with only one fitted density curve. A hanging histogram aligns the
tops of the histogram bars (displayed as lines) with the fitted curve. The lines are positioned at the
midpoints of the histogram bins. A hanging histogram is a goodness-of-fit diagnostic in the sense that
the closer the lines are to the horizontal axis, the better the fit. Hanging histograms are discussed by
Tukey (1977), Wainer (1974), and Velleman and Hoaglin (1981).

displays a fitted inverse Gaussian density curve on the histogram. The curve equation is
( 1=2

p.x/ D hv 2x 3
exp. 22 x
.x /2 / for x > 0
0 for x 0


and
8
:
You can specify values for and with the MU= and LAMBDA= iGauss-options. By default, the
sample mean is used for and a maximum likelihood estimate is computed for .
The IGAUSS option can appear only once in a HISTOGRAM statement. Table 6.24 lists secondary
options you can specify with the IGAUSS option. See “Formulas for Fitted Curves” on page 349.
INDICES
requests capability indices based on the fitted distribution. Enclose the keyword INDICES in parenthe-
ses after the distribution keyword. See “Indices Using Fitted Curves” on page 367 for computational
details and see Output 6.11.2.
K=NORMAL | QUADRATIC | TRIANGULAR

specifies the kernel function (normal, quadratic, or triangular) used to compute a kernel density estimate.
Enclose the K= option in parentheses after the KERNEL option, as in the following statements:
proc capability;
histogram length / kernel(k=quadratic);
run;
You can specify kernel functions for up to five estimates. You can also use the K= option together with
the C= option, which specifies standardized bandwidths. If you specify more kernel functions than
bandwidths, the last bandwidth in the list is repeated for the remaining estimates. Likewise, if you
specify more bandwidths than kernel functions, the last kernel function is repeated for the remaining
estimates. For example, the following statements compute three estimates with bandwidths of 0.5, 1.0,
and 1.5:
proc capability;
histogram length / kernel(c=0.5 1.0 1.5 k=normal quadratic);
run;
The first estimate uses a normal kernel, and the last two estimates use a quadratic kernel. By default, a
normal kernel is used.
KERNEL< ( kernel-options ) >

superimposes up to five kernel density estimates on the histogram. You can specify the kernel-options
described in the following table:
Option Description
C= Specifies the smoothing parameter
COLOR= Specifies the color of the curve
FILL Specifies that the area under the curve is to be filled
K= Specifies the type of kernel function
L= Specifies the line style for the curve
LOWER= Specifies the lower bound for the curve
SYMBOL= Specifies the character used for the kernel density curve in line
printer plots
UPPER= Specifies the upper bound for the curve
W= Specifies the width of the curve
You can request multiple kernel density estimates on the same histogram by specifying a list of values
for either the C= or K= option. For more information, see the entries for these options. Also see
Output 6.6.1 and “Kernel Density Estimates” on page 360. By default, kernel density estimates are
computed using the AMISE method.
LAMBDA=value
specifies the shape parameter for fitted curves requested with the IGAUSS option. Enclose the
LAMBDA= option in parentheses after the IGAUSS distribution keyword. If you do not specify a
displays a fitted lognormal density curve on the histogram. The curve equation is

.log.x / /2
(
p hv exp for x >
p.x/ D 2.x / 2 2
0 for x
where
D scale parameter
and
8
:
Note that the lognormal distribution is also referred to as the SL distribution in the Johnson system of
distributions.
The parameter for the lognormal distribution must be less than the minimum data value. You
can specify with the THETA= lognormal-option. The default value for is zero. If you specify
THETA=EST, a maximum likelihood estimate is computed for . You can specify the parameters
and with the SIGMA= and ZETA= lognormal-options. By default, estimates of and are computed
as described in “Lognormal Distribution” on page 353. For example, the following statements fit a
lognormal distribution function with D 0 and estimates for and :
proc capability;
histogram length / lognormal;
run;
The LOGNORMAL option can appear only once in a HISTOGRAM statement. Table 6.24 lists
secondary options that you can specify with the LOGNORMAL option. See Example 6.9 and
“Formulas for Fitted Curves” on page 349.
LOWER=value-list
specifies lower bounds for kernel density estimates requested with the KERNEL option. Enclose the
LOWER= option in parentheses after the KERNEL option. You can specify up to five lower bounds
for multiple kernel density estimates. If you specify more kernel estimates than lower bounds, the last
lower bound is repeated for the remaining estimates.
MAXNBIN=n
specifies the maximum number of bins to be displayed in a comparative histogram. This option is
useful in situations where the scales or ranges of the data distributions differ greatly from cell to cell.
By default, the bin size and midpoints are determined for the key cell, and then the midpoint list is
extended to accommodate the data ranges for the remaining cells. However, if the cell scales differ
considerably, the resulting number of bins may be so great that each cell histogram is scaled into a
narrow region. By limiting the number of bins with the MAXNBIN= option, you can narrow the
window about the data distribution in the key cell. Note that the MAXNBIN= option provides an
alternative to the MAXSIGMAS= option.
MAXSIGMAS=value
limits the number of bins to be displayed to a range of value standard deviations (of the data in the key
cell) above and below the mean of the data in the key cell. This option is useful in situations where the
data ranges for the remaining cells. If the cell scales differ considerably, however, the resulting number
of bins with the MAXSIGMAS= option, you narrow the window about the data distribution in the key
cell. Note that the MAXSIGMAS= option provides an alternative to the MAXNBIN= option.
MIDPERCENTS
requests a table listing the midpoints and percent of observations in each histogram interval. For
example, the following statements create the table in Figure 6.13:
proc capability;
histogram Length / midpercents;
run;
Figure 6.13 Table of Midpoints and Observed Percentages

Histogram Bins for

Length
Bin Observed
Midpoint Percent
10.02 12.000
10.08 32.000
10.14 28.000
10.20 18.000
10.26 6.000
10.32 4.000
If you specify the MIDPERCENTS option in parentheses after a density estimate option, a table listing
the midpoints, observed percent of observations, and the estimated percent of the population in each
interval (estimated from the fitted distribution) is printed.
The following statements create the table shown in Figure 6.14:
proc capability;
histogram Length / gamma(theta=3 midpercents);
run;
Figure 6.14 Table of Observed and Expected Percentages

Fitted Gamma Distribution for Length (Attachment Point Offset in mm)
Histogram Bin Percents for

Gamma Distribution
Percent
Bin
Midpoint Observed Estimated
10.02 12.000 11.480
10.08 32.000 26.182
10.14 28.000 31.354
10.20 18.000 19.916
10.26 6.000 6.766
10.32 4.000 1.238
MIDPOINTS=value-list | KEY | UNIFORM

specifies how to determine the midpoints for the histogram intervals, where values-list determines the
width of the histogram bars as the difference between consecutive midpoints. The procedure uses the
same values for all variables. See Output 6.9.1.
The range of midpoints, extended at each end by half of the bar width, must cover the range of the data
as well as any specification limits. For example, if you specify
midpoints=2 to 10 by 0.5
then all of the observations and specification limits should fall between 1.75 and 10.25. (Otherwise, a
default list of midpoints is used.) You must use evenly spaced midpoints listed in increasing order.
KEY determines the midpoints for the data in the key cell. The initial number of midpoints
is based on the number of observations in the key cell that use the method of Terrell
and Scott (1985). The procedure extends the midpoint list for the key cell in either
direction as necessary until it spans the data in the remaining cells.
UNIFORM determines the midpoints by using all the observations as if there were no cells. In
other words, the number of midpoints is based on the total sample size by using the
method of Terrell and Scott (1985).
Neither KEY nor UNIFORM apply unless you use the CLASS statement. By default, if you
use a CLASS statement, MIDPOINTS=KEY. However, if the key cell is empty then MID-
POINTS=UNIFORM. Otherwise, the procedure computes the midpoints by using the algorithm
described in Terrell and Scott (1985). The default midpoints are primarily applicable to continuous
data that are approximately normally distributed.
If you produce traditional graphics and use the MIDPOINTS= and HAXIS= options, you can use the
ORDER= option in the AXIS statement you specified with the HAXIS= option. However, for the tick
mark labels to coincide with the histogram interval midpoints, the range of the ORDER= list must
encompass the range of the MIDPOINTS= list, as illustrated in the following statements:
proc capability;
histogram length / midpoints=20 to 80 by 10
haxis=axis1;
axis1 length=6 in order=10 20 30 40 50 60 70 80 90;
run;
MIDPTAXIS=name
is an alias for the HAXIS= option.
MU=value-list
specifies the parameter for fitted curves requested with the GUMBEL, IGAUSS, and NORMAL
options. Enclose the MU= option in parentheses after the distribution keyword. For the normal and
inverse Gaussian distributions, the default value of is the sample mean. If you do not specify a value
for for the Gumbel distribution, the procedure calculates a maximum likelihood estimate.
NENDPOINTS=n
specifies the number of histogram interval endpoints and causes the endpoints, rather than interval
midpoints, to be aligned with horizontal axis tick marks.
NMIDPOINTS=n
specifies the number of histogram intervals.
NOBARS
suppresses drawing of histogram bars. This option is useful when you want to display fitted curves
only.
NOCURVELEGEND
NOCURVEL
suppresses the portion of the legend for fitted curves. If you use the INSET statement to display
information about the fitted curve on the histogram, you can use the NOCURVELEGEND option to
prevent the information about the fitted curve from being repeated in a legend at the bottom of the
histogram. See Output 6.15.1.
NOLEGEND
suppresses legends for specification limits, fitted curves, and hidden observations. See Example 6.13.
Specifying the NOLEGEND option is equivalent to specifying LEGEND=NONE.
NOPLOT
suppresses the creation of a plot. Use the NOPLOT option when you want only to print summary
statistics for a fitted density or create either an OUTFIT= or an OUTHISTOGRAM= data set. See
Example 6.11.
NOPRINT
suppresses printed output summarizing the fitted curve. Enclose the NOPRINT option in parentheses
following the distribution option. See “Customizing a Histogram” on page 316 for an example.

displays a fitted normal density curve on the histogram. The curve equation is
hv
exp 12 . x /2 for 1 < x < 1

p.x/ D p 2
where
D mean
and
8
:
Note that the normal distribution is also referred to as the SN distribution in the Johnson system of
distributions.
You can specify values for and with the MU= and SIGMA= normal-options, as shown in the
following statements:
proc capability;
histogram length / normal(mu=14 sigma=0.05);
run;
By default, the sample mean and sample standard deviation are used for and . The NORMAL
option can appear only once in a HISTOGRAM statement. Table 6.24 lists secondary options that you
can specify with the NORMAL option. See Figure 6.10 and “Formulas for Fitted Curves” on page 349.
NOSPECLEGEND
NOSPECL
suppresses the portion of the legend for specification limit reference lines. See Figure 6.11.
NOTABCONTENTS
suppresses the table of contents entries for tables produced by the HISTOGRAM statement. See
the section “ODS Tables” on page 372 for descriptions of the tables produced by the HISTOGRAM
statement.
OPTBOUNDRANGE=value
defines the sampling range for each parameter during maximum likelihood estimation for the Johnson
SU distribution. PROC CAPABILITY computes initial estimates for each parameter by using the
method of percentiles. The value determines the range of parameter values around the initial estimate
that can be sampled for local optimization starting values. The default is 100.
OPTMAXITER=value
limits the number of iterations that are used by the optimizer in maximum likelihood estimation for the
Johnson SU distribution. The default is 500.
OPTMAXSTARTS=N
defines the maximum number of starting points to be used for local optimization in maximum likelihood
estimation for the Johnson SU distribution. That is, no more than N local optimizations are used in the
multistart algorithm. The default value is 100.
OPTPRINT
prints the iteration history for the Johnson SU distribution maximum likelihood estimation.
OPTSEED=value
specifies a positive integer seed for generating random number sequences in Johnson SU distribution
maximum likelihood estimation. You can use this option to replicate results from different runs.
OPTTOLERANCE=value
specifies the tolerance for declaring optimality in maximum likelihood estimation for the Johnson SU
distribution. The default value is 1E–8.
OUTFIT=SAS-data-set
creates a SAS data set that contains parameter estimates for fitted curves and related goodness-of-fit
information. See “Output Data Sets” on page 369.
OUTHISTOGRAM=SAS-data-set
OUTHIST=SAS-data-set
creates a SAS data set that contains information about histogram intervals. Specifically, the data set
contains the midpoints of the histogram intervals, the observed percent of observations in each interval,
and the estimated percent of observations in each interval (estimated from each of the specified fitted
curves). See “Output Data Sets” on page 369.
OUTKERNEL=SAS-data-set
creates a SAS data set containing information about kernel density estimates requested with the
KERNEL option. See “OUTKERNEL= Output Data Set” on page 372 for details.

displays a fitted generalized Pareto density curve on the histogram. The curve equation is
(
hv
.1 ˛.x /= /1=˛ 1 if ˛ ¤ 0
p.x/ D hv
exp. .x /=/ if ˛ D 0
where
and
8
:
Pareto-option. The default value for is zero. If you specify THETA=EST, a maximum likelihood
estimate is computed for . In addition, the generalized Pareto distribution has a shape parameter ˛ and
a scale parameter . You can specify these parameters with the ALPHA= and SIGMA= Pareto-options.
By default, maximum likelihood estimates are computed for ˛ and .
The PARETO option can appear only once in a HISTOGRAM statement. Table 6.24 lists secondary
options you can specify with the PARETO option. See “Formulas for Fitted Curves” on page 349.
PCTAXIS=name|value-list
is an alias for the VAXIS= option.
PERCENTS=value-list
PERCENT=value-list
specifies a list of percents for which quantiles calculated from the data and quantiles estimated from
the fitted curve are tabulated. The percents must be between 0 and 100. Enclose the PERCENTS=
option in parentheses after the curve option. The default percents are 1, 5, 10, 25, 50, 75, 90, 95, and
99.
For example, the following statements create the table shown in Figure 6.15:
proc capability;
histogram Length / lognormal(percents=1 3 5 95 97 99);
run;
Figure 6.15 Estimated and Observed Quantiles for the Lognormal Curve
Fitted Lognormal Distribution for Length (Attachment Point Offset in mm)
Quantiles for Lognormal

Distribution
Quantile
1.0 10.0180 9.95696
3.0 10.0180 9.98937
5.0 10.0310 10.00658
95.0 10.2780 10.24963
97.0 10.2930 10.26729
99.0 10.3220 10.30071

displays a fitted power function density curve on the histogram. The curve equation is
( ˛ 1
hv ˛ x for < x < C
p.x/ D
0 for x or x C
where
and
8
:
The parameter must be less than or equal to the minimum data value. You can specify and with
the THETA= and the SIGMA= power-options. The default values for and are 0 and 1, respectively.
You can specify THETA=EST and SIGMA=EST to request maximum likelihood estimates for and
.
In addition, the generalized Pareto distribution has a shape parameter ˛. You can specify ˛ with the
ALPHA= power-option. By default, a maximum likelihood estimate is computed for ˛.
The POWER option can appear only once in a HISTOGRAM statement. Table 6.24 lists secondary
options you can specify with the POWER option. See “Formulas for Fitted Curves” on page 349.

displays a fitted Rayleigh density curve on the histogram. The curve equation is
(
2 2
hv x 2 e .x / =.2 / for x
p.x/ D
0 for x <
where
and
8
:
THETA= Rayleigh-option. The default value for is zero. If you specify THETA=EST, a maximum
likelihood estimate is computed for . You can specify with the SIGMA= Rayleigh-option. By
default, a maximum likelihood estimate is computed for .
The RAYLEIGH option can appear only once in a HISTOGRAM statement. Table 6.24 lists secondary
options you can specify with the RAYLEIGH option. See “Formulas for Fitted Curves” on page 349.
RTINCLUDE
includes the right endpoint of each histogram interval in that interval. By default, the left endpoint is
included in the histogram interval.
SB< (SB -options ) >

displays a fitted Johnson SB density curve on the histogram. The curve equation is
8 h i 1
ıhv x x
ˆ
ˆ p
ˆ 2 1
< 2
p.x/ D exp 1 x
C ı log. C / for < x < C
ˆ
ˆ 2 x
ˆ
0 for x or x C
:
where
D threshold parameter . 1 < < 1/

ı D shape parameter .ı > 0/
D shape parameter . 1 < < 1/
and
8
:
The SB distribution is bounded below by the parameter and above by the value C . The parameter
must be less than the minimum data value. You can specify with the THETA= SB -option, or you
can request that be estimated with the THETA = EST SB -option. The default value for is zero.
The sum C must be greater than the maximum data value. The default value for is one. You
can specify with the SIGMA= SB -option, or you can request that be estimated with the SIGMA
= EST SB -option. You can specify ı with the DELTA= SB -option, and you can specify with the
GAMMA= SB -option. Note that the SB -options are given in parentheses after the SB option.
By default, the method of percentiles is used to estimate the parameters of the SB distribution.
Alternatively, you can request the method of moments or the method of maximum likelihood with the
FITMETHOD = MOMENTS or FITMETHOD = MLE options, respectively. Consider the following
example:
proc capability;
histogram length / sb;
histogram length / sb( theta=est sigma=est );
histogram length / sb( theta=0.5 sigma=8.4
delta=0.8 gamma=-0.6 );
run;
The first HISTOGRAM statement fits an SB distribution with default values of D 0 and D 1 and
with percentile-based estimates for ı and . The second HISTOGRAM statement estimates all four
parameters with the method of percentiles. The third HISTOGRAM statement displays an SB curve
with specified values for all four parameters.
The SB option can appear only once in a HISTOGRAM statement. Table 6.24 lists secondary options
you can specify with the SB option.
SIGMA=value-list
specifies the parameter for fitted curves requested with the BETA, EXPONENTIAL, GAMMA,
GUMBEL, LOGNORMAL, NORMAL, PARETO, POWER, RAYLEIGH, SB, SU, and WEIBULL
options. Enclose the SIGMA= option in parentheses after the distribution keyword. Table 6.28
summarizes the use of the SIGMA= option.

BETA Scale parameter 1 SCALE=
EXPONENTIAL Scale parameter Maximum likelihood estimate SCALE=
GAMMA Scale parameter Maximum likelihood estimate SCALE=
GUMBEL Scale parameter Maximum likelihood estimate
LOGNORMAL Shape parameter Estimate computed as described SHAPE=
in “Lognormal Distribution” on
page 353
NORMAL Scale parameter Standard deviation
PARETO Scale parameter Maximum likelihood estimate
POWER Scale parameter 1 SCALE=
RAYLEIGH Scale parameter Maximum likelihood estimate
SB Scale parameter 1 SCALE=

SU Scale parameter Percentile-based estimate SCALE=
WEIBULL Scale parameter Maximum likelihood estimate SCALE=
If you specify SIGMA=EST, an estimate is computed for . For syntax examples, see the entries for
the distribution options.
SU< (SU -options ) >

displays a fitted Johnson SU density curve on the histogram. The curve equation is
8
ıhv p 1
ˆ p
2 1C..x /=/2
ˆ
ˆ
< 2
p.x/ D exp 1
C ı sinh 1 x
for x >
ˆ
ˆ 2
ˆ
0 for x
:
where
D location parameter . 1 < < 1/

and
8
:
You can specify the parameters with the THETA=, SIGMA=, DELTA=, and GAMMA= SU -options,
which are enclosed in parentheses after the SU option. If you do not specify these parameters, they are
estimated.
By default, the method of percentiles is used to estimate the parameters of the SU distribution.
Alternatively, you can request the method of moments or the method of maximum likelihood with the
FITMETHOD = MOMENTS or FITMETHOD = MLE options, respectively. Consider the following
example:
proc capability;
histogram length / su;
histogram length / su( theta=0.5 sigma=8.4
delta=0.8 gamma=-0.6 );
run;
The first HISTOGRAM statement estimates all four parameters with the method of percentiles. The
second HISTOGRAM statement displays an SU curve with specified values for all four parameters.
The SU option can appear only once in a HISTOGRAM statement. Table 6.24 lists secondary options
you can specify with the SU option.
THETA=value-list
THRESHOLD=value-list
specifies the lower threshold parameter for curves requested with the BETA, EXPONENTIAL,
GAMMA, LOGNORMAL, PARETO, POWER, RAYLEIGH, SB, and WEIBULL options, and the
location parameter for curves requested with the SU option. Enclose the THETA= option in
parentheses after the curve option. See Example 6.8. The default value is zero. If you specify
THETA=EST, an estimate is computed for .
UPPER=value-list
specifies upper bounds for kernel density estimates requested with the KERNEL option. Enclose the
UPPER= option in parentheses after the KERNEL option. You can specify up to five upper bounds for
multiple kernel density estimates. If you specify more kernel estimates than upper bounds, the last
upper bound is repeated for the remaining estimates.
VSCALE=COUNT | PERCENT | PROPORTION

specifies the scale of the vertical axis. The value COUNT scales the data in units of the number of
observations per data unit. The value PERCENT scales the data in units of percent of observations per
data unit. The value PROPORTION scales the data in units of proportion of observations per data unit.
See Figure 6.11 for an illustration of VSCALE=COUNT. The default is PERCENT.

displays a fitted Weibull density curve on the histogram. The curve equation is
chv x c 1
. / exp. . x /c / for x >
p.x/ D
0 for x
where
and
8
:
Weibull-option. The default value for is zero. If you specify THETA=EST, a maximum likelihood
estimate is computed for . You can specify and c with the SIGMA= and C= Weibull-options.
By default, maximum likelihood estimates are computed for c and . For example, the following
statements fit a Weibull distribution with D 15 and with maximum likelihood estimates for and c:
proc capability;
histogram length / weibull(theta=15);
run;
Note that the maximum likelihood estimate of c is calculated iteratively using the Newton-Raphson
approximation. The CDELTA=, CINITIAL=, and MAXITER= Weibull-options control the approxima-
tion.
The WEIBULL option can appear only once in a HISTOGRAM statement. Table 6.24 lists secondary
options that you can specify with the WEIBULL option. See Example 6.9 and “Formulas for Fitted
Curves” on page 349.
ZETA=value-list
specifies a value for the scale parameter for lognormal density curves requested with the LOGNOR-
MAL option. Enclose the ZETA= option in parentheses after the LOGNORMAL option. By default,
the procedure calculates a maximum likelihood estimate for . You can specify the SCALE= option as
an alias for the ZETA= option.

BARWIDTH=value
specifies the width of the histogram bars in screen percent units.
BMCFILL=color
specifies the fill color for a box-and-whisker plot in a bottom margin requested with the BMPLOT=
option. By default, the box-and-whisker plot is not filled.
BMCFRAME=color
specifies the color for filling the frame of a bottom margin plot requested with the BMPLOT= option.
By default, this area is not filled.
BMCOLOR=color
specifies the color of a carpet plot, or the outline color of a box-and-whisker plot, in a bottom margin
plot requested with the BMPLOT= option.
BMMARGIN=height
specifies the height in screen percentage units of a bottom margin plot requested with the BMPLOT=
option. By default, a bottom margin plot occupies 15 percent of the vertical display space.
CBARLINE=color
specifies the color of the outline of histogram bars. This option overrides the C= option in the
SYMBOL1 statement.
CFILL=color
specifies a color used to fill the bars of the histogram (or the area under a fitted curve if you also
specify the FILL option). See the entries for the FILL and PFILL= options for additional details. See
Figure 6.11 and Output 6.8.1. Refer to SAS/GRAPH: Reference for a list of colors. By default, bars are
filled with an appropriate color from the ODS style.
CGRID=color
specifies the color for grid lines requested with the GRID option. By default, grid lines are the same
color as the axes. If you use CGRID=, you do not need to specify the GRID option.
CLIPREF
draws reference lines requested with the HREF= and VREF= options behind the histogram bars. By
default, reference lines are drawn in front of the histogram bars.
CLIPSPEC=CLIP | NOFILL
specifies that histogram bars are clipped at the upper and lower specification limit lines when there are
no observations outside the specification limits. The bar intersecting the lower specification limit is
clipped if there are no observations less than the lower limit; the bar intersecting the upper specification
limit is clipped if there are no observations greater than the upper limit. If you specify CLIPSPEC=CLIP,
the histogram bar is truncated at the specification limit. If you specify CLIPSPEC=NOFILL, the portion
of a filled histogram bar outside the specification limit is left unfilled. Specifying CLIPSPEC=NOFILL
when histogram bars are not filled has no effect.
CURVELEGEND=name | NONE
specifies the name of a LEGEND statement describing the legend for specification limits and fitted
curves. Specifying CURVELEGEND=NONE suppresses the legend for fitted curves; this is equivalent
to specifying the NOCURVELEGEND option.
FRONTREF
draws reference lines requested with the HREF= and VREF= options in front of the histogram bars.
When the NOGSTYLE system option is specified, reference lines are drawn behind the histogram bars
by default, and can be obscured by them.
HOFFSET=value
specifies the offset in percent screen units at both ends of the horizontal axis. Specify HOFFSET=0 to
eliminate the default offset.
INTERBAR=value
specifies the horizontal space in percent screen units between histogram bars. By default, the bars are
contiguous.
LEGEND=name | NONE
and fitted curves. Specifying LEGEND=NONE suppresses all legend information and is equivalent to
specifying the NOLEGEND option.
LGRID=n
specifies the line type for the grid requested with the GRID option. If you use the LGRID= option, you
do not need to specify the GRID option. The default is 1, which produces a solid line.
PFILL=pattern
specifies a pattern used to fill the bars of the histograms (or the areas under a fitted curve if you also
specify the FILL option). See the entries for the CFILL= and FILL options for additional details. Refer
to SAS/GRAPH: Reference for a list of pattern values. By default, the bars and curve areas are not
filled.
Details: HISTOGRAM Statement F 349
SPECLEGEND=name | NONE
specifies the name of a LEGEND statement describing the legend for specification limits and fitted
curves. Specifying SPECLEGEND=NONE, which suppresses the portion of the legend for specification
limit references lines, is equivalent to specifying the NOSPECLEGEND option.
VOFFSET=value
specifies the offset in percent screen units at the upper end of the vertical axis.
WBARLINE=n
specifies the width of bar outlines. By default, n = 1.
WGRID=n
specifies the width of the grid lines requested with the GRID option. By default, grid lines are the
same width as the axes. If you use the WGRID= option, you do not need to specify the GRID option.
Options for Legacy Line Printer Charts

specifies the character used for the density curve or kernel density curve in line printer plots. Enclose
the SYMBOL= option in parentheses after the distribution option or the KERNEL option. The default
character is the first letter of the distribution keyword or ‘1’ for the first kernel density estimate,
‘2’ for the second kernel density estimate, and so on. If you use the SYMBOL= option with the
KERNEL option, you can specify a list of up to five characters in parentheses for multiple kernel
density estimates. If there are more estimates than characters, the last character specified is used for
the remaining estimates.
Details: HISTOGRAM Statement

formulas for fitted distributions
formulas for kernel density estimates
printed output
OUTFIT=, OUTHISTOGRAM=, and OUTKERNEL= data sets
graphical enhancements to histograms
Formulas for Fitted Curves

The following sections provide information about the families of parametric distributions that you can fit
with the HISTOGRAM statement. Properties of these distributions are discussed by Johnson, Kotz, and
Balakrishnan (1994) and Johnson, Kotz, and Balakrishnan (1995).
Beta Distribution
The fitted density function is
( ˛ 1
.x / . C x/ˇ 1
B.˛;ˇ / .˛Cˇ 1/
hv for < x < C
p.x/ D
0 for x or x C
.˛/.ˇ /
and

v D vertical scaling factor, and
8
:
N OTE : This notation is consistent with that of other distributions that you can fit with the HISTOGRAM
statement. However, many texts, including Johnson, Kotz, and Balakrishnan (1995), write the beta density
function as
(
.x a/p 1 .b x/q 1
B.p;q/.b a/pCq 1
for a < x < b
p.x/ D
0 for x a or x b
The two notations are related as follows:
Db a
Da
˛Dp
ˇDq
The range of the beta distribution is bounded below by a threshold parameter D a and above by C D b.
If you specify a fitted beta curve by using the BETA option, must be less than the minimum data value,
and C must be greater than the maximum data value. You can specify and with the THETA= and
SIGMA= beta-options in parentheses after the keyword BETA. By default, D 1 and D 0. If you specify
THETA=EST and SIGMA=EST, maximum likelihood estimates are computed for and .
In addition, you can specify ˛ and ˇ with the ALPHA= and BETA= beta-options, respectively. By default,
the procedure calculates maximum likelihood estimates for ˛ and ˇ. For example, to fit a beta density curve
to a set of data bounded below by 32 and above by 212 with maximum likelihood estimates for ˛ and ˇ, use
the following statement:
histogram length / beta(theta=32 sigma=180);

The beta distributions are also referred to as Pearson Type I or II distributions. These include the power-
function distribution (ˇ D 1), the arc-sine distribution (˛ D ˇ D 12 ), and the generalized arc-sine
distributions (˛ C ˇ D 1, ˇ ¤ 12 ).
You can use the DATA step function BETAINV to compute beta quantiles and the DATA step function
PROBBETA to compute beta probabilities.
Exponential Distribution
hv x
p.x/ D exp. . // for x
0 for x <
where
8
:
The threshold parameter must be less than or equal to the minimum data value. You can specify with the
THRESHOLD= exponential-option. By default, D 0. If you specify THETA=EST, a maximum likelihood
estimate is computed for . In addition, you can specify with the SCALE= exponential-option. By default,
the procedure calculates a maximum likelihood estimate for . Note that some authors define the scale
parameter as 1 .
The exponential distribution is a special case of both the gamma distribution (with ˛ D 1) and the Weibull
distribution (with c = 1). A related distribution is the extreme value distribution. If Y D exp. X / has an
exponential distribution, then X has an extreme value distribution.
Gamma Distribution
(
hv
.˛/
. x /˛ 1 exp. . x // for x >
p.x/ D
0 for x
where
8
:
The threshold parameter must be less than the minimum data value. You can specify with the THRESH-
OLD= gamma-option. By default, D 0. If you specify THETA=EST, a maximum likelihood estimate is
computed for . In addition, you can specify and ˛ with the SCALE= and ALPHA= gamma-options. By
default, the procedure calculates maximum likelihood estimates for and ˛.
The gamma distributions are also referred to as Pearson Type III distributions, and they include the chi-square,
exponential, and Erlang distributions. The probability density function for the chi-square distribution is
(
1

x 2 1

2. 2 / 2
exp. x2 / for x > 0
p.x/ D
0 for x 0
Notice that this is a gamma distribution with ˛ D 2 , D 2, and D 0. The exponential distribution
is a gamma distribution with ˛ D 1, and the Erlang distribution is a gamma distribution with ˛ being a
1 ;:::;Xn /
positive integer. A related distribution is the Rayleigh distribution. If R D max.X
min.X1 ;:::;Xn /
where the Xi ’s
2
are independent variables, then log R is distributed with a distribution having a probability density
function of
( h

i 1 2
2 2 1 . 2 / x 1 exp. x2 / for x > 0
p.x/ D
0 for x 0
If D 2, the preceding distribution is referred to as the Rayleigh distribution.

You can use the DATA step function GAMINV to compute gamma quantiles and the DATA step function
PROBGAM to compute gamma probabilities.
Gumbel Distribution
hv .x /=

.x /=

p.x/ D e exp e

where
8
:
You can specify and with the MU= and SIGMA= Gumbel-options, respectively. By default, the procedure
calculates maximum likelihood estimates for these parameters.
N OTE : The Gumbel distribution is also referred to as Type 1 extreme value distribution.
N OTE : The random variable X has Gumbel (Type 1 extreme value) distribution if and only if e X has Weibull
distribution and exp..X /= / has standard exponential distribution.
Inverse Gaussian Distribution

( 1=2

p.x/ D hv 2x 3 exp 22 x
.x /2 for x > 0
0 for x 0
where
D location parameter . > 0/

8
:
The location parameter has to be greater then zero. You can specify with the MU= iGauss-option. In
addition, you can specify shape parameter with the LAMBDA= iGauss-option. By default, the procedure
uses the sample mean for and calculates a maximum likelihood estimate for .
N OTE : The special case where D 1 and D corresponds to the Wald distribution.
Lognormal Distribution

.log.x / /2
(
p hv exp for x >
p.x/ D 2.x / 2 2
0 for x
where
D scale parameter . 1 < < 1/
8
:
OLD= lognormal-option. By default, D 0. If you specify THETA=EST, a maximum likelihood estimate is
computed for . You can specify and with the SCALE= and SHAPE= lognormal-options, respectively.
By default, the procedure calculates an estimate of as
Pn
O D i D1 log .xi /
n
and an estimate of as
s
Pn
i D1 .log .xi / /2
O D
n 1
N OTE : The lognormal distribution is also referred to as the SL distribution in the Johnson system of
distributions.
N OTE : This book uses to denote the shape parameter of the lognormal distribution, whereas is used to
denote the scale parameter of the beta, exponential, gamma, Gumbel, inverse Gaussian, normal, generalized
Pareto, power function, Rayleigh, and Weibull distributions. The use of to denote the lognormal shape
parameter is based on the fact that 1 .log.X / / has a standard normal distribution if X is lognormally
distributed.
Normal Distribution
hv 1 x 2

p.x/ D p exp 2. / for 1 < x < 1
2
where
D mean
8
:
You can specify and with the MU= and SIGMA= normal-options, respectively. By default, the procedure
estimates with the sample mean and with the sample standard deviation.
You can use the DATA step function PROBIT to compute normal quantiles and the DATA step function
PROBNORM to compute probabilities.
N OTE : The normal distribution is also referred to as the SN distribution in the Johnson system of distributions.
Generalized Pareto Distribution

(
hv
.1 ˛.x /= /1=˛ 1 if ˛ ¤ 0
p.x/ D hv
exp. .x /= / if ˛ D 0
where

8
:
The support of the distribution is x > for ˛ 0 and < x < =˛ for ˛ > 0.
N OTE : Special cases of the generalized Pareto distribution with ˛ D 0 and ˛ D 1 correspond respectively to
the exponential distribution with mean and uniform distribution on the interval .; /.
The threshold parameter must be less than the minimum data value. You can specify with the THETA=
Pareto-option. By default, D 0. You can also specify ˛ and with the ALPHA= and SIGMA= Pareto-
options,respectively. By default, the procedure calculates maximum likelihood estimates for these parameters.
N OTE : Maximum likelihood estimation of the parameters works well if ˛ < 12 , but not otherwise. In
this case the estimators are asymptotically normal and asymptotically efficient. The asymptotic normal
distribution of the maximum likelihood estimates has mean .˛; / and variance-covariance matrix
.1 ˛/2

1 .1 ˛/
:
n .1 ˛/ 2 2 .1 ˛/
N OTE : If no local minimum is found in the region
f˛ < 0; > 0g [ f0 < ˛ 1; =˛ > max.Xi /g;
there is no maximum likelihood estimator. More details on how to find maximum likelihood estimators and a
suggested algorithm can be found in Grimshaw (1993).
Power Function Distribution

( ˛ 1
hv ˛ x for < x < C
p.x/ D
0 for x or x C
where

8
:
N OTE : This notation is consistent with that of other distributions that you can fit with the HISTOGRAM
statement. However, many texts, including Johnson, Kotz, and Balakrishnan (1995), write the density
function of power function distribution as
(
p x a p 1

for a < x < b
p.x/ D b a b a
0 for x a or x b
The two parameterizations are related as follows:
Db a
Da
˛Dp
N OTE : The family of power function distributions is a subclass of beta distribution with density function
(
/˛ 1 .C x/ˇ 1
hv .x B.˛;ˇ / .˛Cˇ 1/
for < x < C
p.x/ D
0 for x or x C
where B.˛; ˇ/ D .˛/.ˇ

.˛Cˇ /
/
with parameter ˇ D 1. Therefore, all properties and estimation procedures of
beta distribution apply.
The range of the power function distribution is bounded below by a threshold parameter D a and above by
C D b. If you specify a fitted power function curve by using the POWER option, must be less than the
minimum data value and C must be greater than the maximum data value. You can specify and with
the THETA= and SIGMA= power-options in parentheses after the keyword POWER. By default, D 1 and
D 0. If you specify THETA=EST and SIGMA=EST, maximum likelihood estimates are computed for
and . However, three-parameter maximum likelihood estimation does not always converge.
In addition, you can specify ˛ with the ALPHA= power-option. By default, the procedure calculates a
maximum likelihood estimate for ˛. For example, to fit a power function density curve to a set of data
bounded below by 32 and above by 212 with maximum likelihood estimate for ˛, use the following statement:
histogram Length / power(theta=32 sigma=180);
Rayleigh Distribution
(
2 2
hv x 2 e .x / =.2 / for x
p.x/ D
0 for x <
where

8
:
N OTE : The Rayleigh distribution is a Weibull distribution with density function

( k 1
k x
p.x/ D hv
exp. . x /k / for x
0 for x <
p
and with shape parameter k = 2 and scale parameter D 2 .
The threshold parameter must be less than the minimum data value. You can specify with the THETA=
Rayleigh-option. By default, D 0. In addition you can specify with the SIGMA= Rayleigh-option. By
default, the procedure calculates maximum likelihood estimate for .
For example, to fit a Rayleigh density curve to a set of data bounded below by 32 with maximum likelihood
estimate for , use the following statement:
histogram Length / rayleigh(theta=32);
Johnson SB Distribution
8 h i 1
ıhv x x
ˆ
ˆ p
1
< 2
ˆ 2
p.x/ D exp 1 x
C ı log. C / for < x < C
ˆ
ˆ 2 x
ˆ
0 for x or x C
:
where
D threshold parameter . 1 < < 1/

8
:
The SB distribution is bounded below by the parameter and above by the value C . The parameter
must be less than the minimum data value. You can specify with the THETA= SB -option, or you can
request that be estimated with the THETA = EST SB -option. The default value for is zero. The sum
C must be greater than the maximum data value. The default value for is one. You can specify with
the SIGMA= SB -option, or you can request that be estimated with the SIGMA = EST SB -option.
By default, the method of percentiles given by Slifker and Shapiro (1980) is used to estimate the parameters.
This method is based on four data percentiles, denoted by x 3z , x z , xz , and x3z , which correspond to the
four equally spaced percentiles of a standard normal distribution, denoted by 3z, z, z, and 3z, under the
transformation

x
z D C ı log
C x
The default value of z is 0.524. The results of the fit are dependent on the choice of z, and you can specify
other values with the FITINTERVAL= option (specified in parentheses after the SB option). If you use the
method of percentiles, you should select a value of z that corresponds to percentiles which are critical to your
application.
The following values are computed from the data percentiles:
m D x3z xz
n D x z x 3z
p D xz x z
It was demonstrated by Slifker and Shapiro (1980) that

mn
p2
> 1 for any SU distribution
mn
p2
< 1 for any SB distribution
mn
p2
D 1 for any SL (lognormal) distribution
A tolerance interval around one is used to discriminate among the three families with this ratio criterion. You
can specify the tolerance with the FITTOLERANCE= option (specified in parentheses after the SB option).
The default tolerance is 0.01. Assuming that the criterion satisfies the inequality
mn
<1 tolerance
p2
the parameters of the SB distribution are computed using the explicit formulas derived by Slifker and Shapiro
(1980).
If you specify FITMETHOD = MOMENTS (in parentheses after the SB option) the method of moments is
used to estimate the parameters. If you specify FITMETHOD = MLE (in parentheses after the SB option) the
method of maximum likelihood is used to estimate the parameters. Note that maximum likelihood estimates
may not always exist. Refer to Bowman and Shenton (1983) for discussion of methods for fitting Johnson
distributions.
Johnson SU Distribution
8
ıhv p 1
ˆ p
2 1C..x /=/2
ˆ
ˆ
< 2
p.x/ D exp 1
C ı sinh 1 x
for x >
ˆ
ˆ 2
ˆ
0 for x
:
where
D location parameter . 1 < < 1/

8
:
You can specify the parameters with the THETA=, SIGMA=, DELTA=, and GAMMA= SU -options, which
are enclosed in parentheses after the SU option. If you do not specify these parameters, they are estimated.
By default, the method of percentiles given by Slifker and Shapiro (1980) is used to estimate the parameters.
This method is based on four data percentiles, denoted by x 3z , x z , xz , and x3z , which correspond to the
four equally spaced percentiles of a standard normal distribution, denoted by 3z, z, z, and 3z, under the
transformation

x
z D C ı sinh 1

The default value of z is 0.524. The results of the fit are dependent on the choice of z, and you can specify
other values with the FITINTERVAL= option (specified in parentheses after the SU option). If you use the
method of percentiles, you should select a value of z that corresponds to percentiles which are critical to your
application.
The following values are computed from the data percentiles:
m D x3z xz
n D x z x 3z
p D xz x z
It was demonstrated by Slifker and Shapiro (1980) that
mn
p2
> 1 for any SU distribution
mn
p2
< 1 for any SB distribution
mn
p2
D 1 for any SL (lognormal) distribution
A tolerance interval around one is used to discriminate among the three families with this ratio criterion. You
can specify the tolerance with the FITTOLERANCE= option (specified in parentheses after the SU option).
The default tolerance is 0.01. Assuming that the criterion satisfies the inequality
mn
> 1 C tolerance
p2
the parameters of the SU distribution are computed using the explicit formulas derived by Slifker and Shapiro
(1980).
If you specify FITMETHOD = MOMENTS (in parentheses after the SU option) the method of moments is
used to estimate the parameters. If you specify FITMETHOD = MLE (in parentheses after the SU option) the
method of maximum likelihood is used to estimate the parameters. Note that maximum likelihood estimates
may not always exist. Refer to Bowman and Shenton (1983) for discussion of methods for fitting Johnson
distributions.
Weibull Distribution
chv x c 1
. / exp. . x /c / for x >
p.x/ D
0 for x
where
8
:
OLD= Weibull-option. By default, D 0. If you specify THETA=EST, a maximum likelihood estimate is
computed for . You can specify and c with the SCALE= and SHAPE= Weibull-options, respectively. By
default, the procedure calculates maximum likelihood estimates for and c.
The exponential distribution is a special case of the Weibull distribution where c = 1.
Kernel Density Estimates

You can use the KERNEL option to superimpose kernel density estimates on histograms. Smoothing the
data distribution with a kernel density estimate can be more effective than using a histogram to examine
features that might be obscured by the choice of histogram bins or sampling variation. A kernel density
estimate can also be more effective than a parametric curve fit when the process distribution is multimodal.
See Example 6.12.
The general form of the kernel density estimator is
n
1 X x x
i
fO .x/ D K0
n
i D1
where K0 ./ is a kernel function, is the bandwidth, n is the sample size, and xi is the ith observation.
The KERNEL option provides three kernel functions (K0 ): normal, quadratic, and triangular. You can specify
the function with the K= kernel-option in parentheses after the KERNEL option. Values for the K= option
are NORMAL, QUADRATIC, and TRIANGULAR (with aliases of N, Q, and T, respectively). By default, a
normal kernel is used. The formulas for the kernel functions are
K0 .t / D p1 1 2
1<t <1
Normal
2
exp. 2t / for
3
Quadratic K0 .t / D 4 .1 t 2/ for jtj 1
Triangular K0 .t / D 1 jtj for jtj 1
The value of , referred to as the bandwidth parameter, determines the degree of smoothness in the estimated
density function. You specify indirectly by specifying a standardized bandwidth c with the C= kernel-option.
If Q is the interquartile range, and n is the sample size, then c is related to by the formula
1
D cQn 5
For a specific kernel function, the discrepancy between the density estimator fO .x/ and the true density f .x/
is measured by the mean integrated square error (MISE):
Z Z
MISE./ D fE.fO .x// f .x/g2 dx C var.fO .x//dx
x x
The MISE is the sum of the integrated squared bias and the variance. An approximate mean integrated square
error (AMISE) is
Z 2 Z
1 1
Z
2
AMISE./ D 4 2
t K.t /dt 00
f .x/ dx C K.t /2 dt
4 t x n t
A bandwidth that minimizes AMISE can be derived by treating f .x/ as the normal density having parameters
and estimated by the sample mean and standard deviation. If you do not specify a bandwidth parameter
or if you specify C=MISE, the bandwidth that minimizes AMISE is used. The value of AMISE can be used
to compare different density estimates. For each estimate, the bandwidth parameter c, the kernel function
type, and the value of AMISE are reported in the SAS log.
The general kernel density estimates assume that the domain of the density to estimate can take on all values
on a real line. However, sometimes the domain of a density is an interval bounded on one or both sides. For
example, if a variable Y is a measurement of only positive values, then the kernel density curve should be
bounded so that it is zero for negative Y values.
The CAPABILITY procedure uses a reflection technique to create the bounded kernel density curve, as
described in Silverman (1986, pp 30-31). It adds the reflections of kernel density that are outside the boundary
to the bounded kernel
estimates. The general form of the bounded kernel density estimator is computed by
replacing K0 x xi in the original equation with

x xi .x xl / C .xi xl / .xu x/ C .xu xi /
K0 C K0 C K0

where xl is the lower bound and xu is the upper bound.

Without a lower bound, xl D 1 and K0 .x xl /C.x
i xl /
equals zero. Similarly, without an upper bound,

xu D 1 and K0 .xu x/C.x
u xi /
equals zero.
When C=MISE is used with a bounded kernel density, the CAPABILITY procedure uses a bandwidth that
minimizes the AMISE for its corresponding unbounded kernel.
Printed Output
If you request a fitted parametric distribution, printed output summarizing the fit is produced in addition to
the graphical display. Figure 6.16 shows the printed output for a fitted lognormal distribution requested by
the following statements:
proc capability data=Hang;

spec target=14 lsl=13.95 usl=14.05;
hist / lognormal(indices midpercents);
run;
Figure 6.16 Sample Summary of Fitted Distribution

Fitted Lognormal Distribution for Width (Width in cm)
Parameters for Lognormal

Distribution
Threshold Theta 0
Scale Zeta 2.638966
Shape Sigma 0.001497
Mean 13.99873
Std Dev 0.020952
Goodness-of-Fit Tests for Lognormal Distribution

Percent Outside Specifications for Lognormal

Distribution
LSL 13.950000 USL 14.050000
Obs Pct < LSL 2.000000 Obs Pct > USL 0
Capability
Indices Based
on Lognormal
Distribution
Cp 0.795463
CPL 0.776822
CPU 0.814021
Cpk 0.776822
Cpm 0.792237
Histogram Bin Percents for

Lognormal Distribution
Percent
Bin
Midpoint Observed Estimated
13.95 4.000 2.963
13.97 18.000 15.354
13.99 26.000 33.872
14.01 38.000 32.055
14.03 10.000 13.050
14.05 4.000 2.281

Distribution
Quantile
1.0 13.9440 13.9501
5.0 13.9656 13.9643
10.0 13.9710 13.9719
25.0 13.9860 13.9846
50.0 14.0018 13.9987
75.0 14.0129 14.0129
90.0 14.0218 14.0256
95.0 14.0241 14.0332
99.0 14.0470 14.0475
The summary is organized into the following parts:
Parameters
Chi-Square Goodness-of-Fit Test
EDF Goodness-of-Fit Tests
Specifications
Indices Using the Fitted Curve
Histogram Intervals
Quantiles
These parts are described in the sections that follow.
Parameters
This section lists the parameters for the fitted curve as well as the estimated mean and estimated standard
deviation. See “Formulas for Fitted Curves” on page 349.
Chi-Square Goodness-of-Fit Test

The chi-square goodness-of-fit statistic for a fitted parametric distribution is computed as follows:
m
X .Oi Ei /2
2 D
Ei
i D1
where
Oi D observed value in ith histogram interval

Ei D expected value in ith histogram interval
m = number of histogram intervals
p = number of estimated parameters
The degrees of freedom for the chi-square test is equal to m p 1. You can save the observed and expected
interval values in the OUTFIT= data set discussed in “Output Data Sets” on page 369.
Note that empty intervals are not combined, and the range of intervals used to compute 2 begins with the
first interval containing observations and ends with the final interval containing observations.
EDF Goodness-of-Fit Tests

When you fit a parametric distribution, the HISTOGRAM statement provides a series of goodness-of-fit
tests based on the empirical distribution function (EDF). The EDF tests offer advantages over the chi-square
goodness-of-fit test, including improved power and invariance with respect to the histogram midpoints. For a
thorough discussion, refer to D’Agostino and Stephens (1986).
The empirical distribution function is defined for a set of n independent observations X1 ; : : : ; Xn with a com-
mon distribution function F .x/. Denote the observations ordered from smallest to largest as X.1/ ; : : : ; X.n/ .
The empirical distribution function, Fn .x/, is defined as
Fn .x/ D 0; x < X.1/

Fn .x/ D ni ; X.i / x < X.iC1/ i D 1; : : : ; n 1
Fn .x/ D 1; X.n/ x
Note that Fn .x/ is a step function that takes a step of height n1 at each observation. This function estimates
the distribution function F .x/. At any value x, Fn .x/ is the proportion of observations less than or equal to x,
while F .x/ is the probability of an observation less than or equal to x. EDF statistics measure the discrepancy
between Fn .x/ and F .x/.
The computational formulas for the EDF statistics make use of the probability integral transformation
U D F .X/. If F .X / is the distribution function of X, the random variable U is uniformly distributed
between 0 and 1.
Given n observations X.1/ ; : : : ; X.n/ , the values U.i / D F .X.i / / are computed by applying the transformation,
as shown in the following sections.
The HISTOGRAM statement provides three EDF tests:
Kolmogorov-Smirnov
Anderson-Darling
Cramér–von Mises
These tests are based on various measures of the discrepancy between the empirical distribution function
Fn .x/ and the proposed parametric cumulative distribution function F .x/.
The following sections provide formal definitions of the EDF statistics.
Kolmogorov-Smirnov Statistic The Kolmogorov-Smirnov statistic (D) is defined as

D D supx jFn .x/ F .x/j
The Kolmogorov-Smirnov statistic belongs to the supremum class of EDF statistics. This class of statistics is
based on the largest vertical difference between F .x/ and Fn .x/.
The Kolmogorov-Smirnov statistic is computed as the maximum of D C and D , where D C is the largest
vertical distance between the EDF and the distribution function when the EDF is greater than the distribution
function, and D is the largest vertical distance when the EDF is less than the distribution function.

D C D maxi ni U.i /

D D maxi U.i / i n1
D max D C ; D

D
Anderson-Darling Statistic The Anderson-Darling statistic and the Cramér–von Mises statistic be-
long to the quadratic class of EDF statistics. This class of statistics is based on the squared difference
.Fn .x/ F .x//2 . Quadratic statistics have the following general form:
Z C1
QDn .Fn .x/ F .x//2 .x/dF .x/
1
The function .x/ weights the squared difference .Fn .x/ F .x//2 .
The Anderson-Darling statistic (A2 ) is defined as
Z C1
2
A Dn .Fn .x/ F .x//2 ŒF .x/ .1 F .x// 1
dF .x/
1
1
Here the weight function is .x/ D ŒF .x/ .1 F .x// .
The Anderson-Darling statistic is computed as
n
2 1 X
A D n .2i 1/ log U.i / C .2n C 1 2i / log f1 U.i /
n
i D1
Cramér–von Mises Statistic The Cramér–von Mises statistic (W 2 ) is defined as

Z C1
W2 Dn .Fn .x/ F .x//2 dF .x/
1
Here the weight function is .x/ D 1.
The Cramér–von Mises statistic is computed as
n
2i 1 2

2
X 1
W D U.i / C
2n 12n
i D1
Probability Values for EDF Tests Once the EDF test statistics are computed, the associated probability
values (p-values) must be calculated.
For the Gumbel, inverse Gaussian, generalized Pareto, and Rayleigh distributions, the procedure computes
associated probability values (p-values) by resampling from the estimated distribution. It generates k random
samples of size n, where k is specified by the EDFNSAMPLES= option and n is the number of observations
in the original data. EDF test statistics are computed for each sample, and the p-value is the proportion of
samples whose EDF statistic is greater than or equal to the statistic computed for the original data. You can
use the EDFSEED= option to specify a seed value for generating the sample values.
For the beta, exponential, gamma, lognormal, normal, power function, and Weibull distributions, the
CAPABILITY procedure uses internal tables of probability levels similar to those given by D’Agostino and
Stephens (1986). If the value is between two probability levels, then linear interpolation is used to estimate
the probability value. The probability value depends upon the parameters that are known and the parameters
that are estimated for the distribution you are fitting. Table 6.29 summarizes different combinations of
estimated parameters for which EDF tests are available.
Table 6.29 Availability of EDF Tests
Distribution Parameters Tests Available

Threshold Scale Shape
Beta known known ˛; ˇ known All
known known ˛; ˇ < 5 unknown All
Exponential known, known All
known unknown All
unknown known All
unknown unknown All
Gamma known known ˛ known All
known unknown ˛ known All
known known ˛ unknown All
known unknown ˛ > 1 unknown All
unknown known ˛ > 1 known All
unknown unknown ˛ > 1 known All
unknown known ˛ > 1 unknown All
unknown unknown ˛ > 1 unknown All
Lognormal known known known All
known known unknown A2 and W 2
known unknown known A2 and W 2
known unknown unknown All
unknown known < 3 known All
unknown known < 3 unknown All
unknown unknown < 3 known All
unknown unknown < 3 unknown All
Normal known known All
known unknown A2 and W 2
unknown known A and W 2
2
unknown unknown All

Distribution Parameters Tests Available

Threshold Scale Shape
Weibull known known c known All
known unknown c known A2 and W 2
known known c unknown A2 and W 2
known unknown c unknown A2 and W 2
unknown known c > 2 known All
unknown unknown c > 2 known All
unknown known c > 2 unknown All
unknown unknown c > 2 unknown All
Specifications
This section is included in the summary only if you provide specification limits, and it tabulates the limits as
well as the observed percentages and estimated percentages outside the limits.
The estimated percentages are computed only if fitted distributions are requested and are based on the
probability that an observed value exceeds the specification limits, assuming the fitted distribution. The
observed percentages are the percents of observations outside the specification limits.
Indices Using Fitted Curves

This section is included in the summary only if you specify the INDICES option in parentheses after a
distribution option, as in the statements that produce Figure 6.16. Standard process capability indices, such
as Cp and Cpk , are not appropriate if the data are not normally distributed. The INDICES option computes
generalizations of the standard indices by using the fact that for the normal distribution, 3 is both the
distance from the lower 0.135 percentile to the median (or mean) and the distance from the median (or
mean) to the upper 99.865 percentile. These percentiles are estimated from the fitted distribution, and the
appropriate percentile-to-median distances are substituted for 3 in the standard formulas.
Writing T for the target, LSL and USL for the lower and upper specification limits, and P˛ for the 100˛th
percentile, the generalized capability indices are as follows:
P0:5 LSL
CPL D
P0:5 P0:00135
USL P0:5
CPU D
P0:99865 P0:5
USL LSL
Cp D
P0:99865 P0:00135

P0:5 LSL USL P0:5
Cpk D min ;
P0:5 P0:00135 P0:99865 P0:5
ˇ1 ˇ
ˇ .USL C LSL/ P0:5 ˇ
2
K D2
USL LSL

T LSL USL T
min P0:5 P0:00135 ; P0:99865 P0:5
Cpm D r 2
T
1C
If the data are normally distributed, these formulas reduce to the formulas for the standard capability indices,
which are given in the section “Standard Capability Indices” on page 246.
The following guidelines apply to the use of generalized capability indices requested with the INDICES
option:
When you choose the family of parametric distributions for the fitted curve, consider whether an
appropriate family can be derived from assumptions about the process.
Whenever possible, examine the data distribution with a histogram, probability plot, or quantile-quantile
plot.
Apply goodness-of-fit tests to assess how well the parametric distribution models the data.
Consider whether a generalized index has a meaningful practical interpretation in your application.
At the time of this writing, there is ongoing research concerning the application of generalized capability
indices, and it is important to note that other approaches can be used with nonnormal data:
Transform the data to normality, then compute and report standard capability indices on the transformed
scale.
Report the proportion of nonconforming output estimated from the fitted distribution.
If it is not possible to adequately model the data distribution with a parametric density, smooth the data
distribution with a kernel density estimate and simply report the proportion of nonconforming output.
Refer to Rodriguez and Bynum (1992) for additional discussion.
Histogram Intervals
This section is included in the summary only if you specify the MIDPERCENTS option in parentheses after
the distribution option, as in the statements that produce Figure 6.16. This table lists the interval midpoints
along with the observed and estimated percentages of the observations that lie in the interval. The estimated
percentages are based on the fitted distribution.
In addition, you can specify the MIDPERCENTS option to request a table of interval midpoints with the
observed percent of observations that lie in the interval. See the entry for the MIDPERCENTS option on
page 336.
Quantiles
This table lists observed and estimated quantiles. You can use the PERCENTS= option to specify the list of
quantiles to appear in this list. The list in Figure 6.16 is the default list. See the entry for the PERCENTS=
option on page 341.
Output Data Sets

You can create two output data sets with the HISTOGRAM statement: the OUTFIT= data set and the
OUTHISTOGRAM= data set. These data sets are described in the following sections.
OUTFIT= Data Set

The OUTFIT= data set contains the parameters of fitted density curves, information about chi-square and EDF
goodness-of-fit tests, specification limit information, and capability indices based on the fitted distribution.
Because you can specify multiple HISTOGRAM statements with the CAPABILITY procedure, you can
create several OUTFIT= data sets. For each variable plotted with the HISTOGRAM statement, the OUTFIT=
data set contains one observation for each fitted distribution requested in the HISTOGRAM statement. If you
use a BY statement, the OUTFIT= data set contains several observations for each BY group (one observation
for each variable and fitted density combination). ID variables are not saved in the OUTFIT= data set.
The OUTFIT= data set contains the variables listed in Table 6.30. By default, an OUTFIT= data set contains
_MIDPT1_ and _MIDPTN_ variables, whose values identify histogram intervals by their midpoints. When
the ENDPOINTS= or NENDPOINTS option is specified, intervals are identified by endpoint values instead.
If the RTINCLUDE option is specified, the variables _MAXPT1_ and _MAXPTN_ contain upper endpoint
values. Otherwise, the variables _MINPT1_ and _MINPTN_ contain lower endpoint values.
Table 6.30 Variables in the OUTFIT= Data Set
_ADASQ_ Anderson-Darling EDF goodness-of-fit statistic
_ADP_ p-value for Anderson-Darling EDF goodness-of-fit test
_CHISQ_ Chi-square goodness-of-fit statistic
_CP_ Generalized capability index Cp based on the fitted curve
_CPK_ Generalized capability index Cpk based on the fitted curve
_CPL_ Generalized capability index CPL based on the fitted curve
_CPM_ Generalized capability index Cpm based on the fitted curve
_CPU_ Generalized capability index CPU based on the fitted curve
_CURVE_ Name of fitted distribution (abbreviated to 8 characters)
_CVMWSQ_ Cramér–von Mises EDF goodness-of-fit statistic
_CVMP_ p-value for Cramér–von Mises EDF goodness-of-fit test
_DF_ Degrees of freedom for chi-square goodness-of-fit test
_ESTGTR_ Estimated percent of population greater than upper specification
limit
_ESTLSS_ Estimated percent of population less than lower specification limit
_ESTSTD_ Estimated standard deviation
_EXPECT_ Estimated mean
_K_ Generalized capability index K based on the fitted curve
_KSD_ Kolmogorov-Smirnov EDF goodness-of-fit statistic
_KSP_ p-value for Kolmogorov-Smirnov EDF goodness-of-fit test
_LOCATN_ Location parameter for fitted distribution. For the Gumbel, inverse
Gaussian, and normal distributions, this is either the value of
specified with the MU= option or the value estimated by the
procedure. For all other distributions, this is either the value
specified or estimated according to the THETA= option, or zero.
_MAXPT1_ Upper endpoint of first interval used to calculate the value of the
chi-square statistic.
_MAXPTN_ Upper endpoint of last interval used to calculate the value of the
_MIDPT1_ Midpoint of first interval used to calculate the value of the
chi-square statistic. This is the leftmost interval that contains at
least one value of the variable.
_MIDPTN_ Midpoint of last interval used to calculate the value of the
chi-square statistic. This is the rightmost interval that contains at
least one value of the variable.
_MINPT1_ Lower endpoint of first interval used to calculate the value of the
_MINPTN_ Lower endpoint of last interval used to calculate the value of the
_OBSGTR_ Observed percent of data greater than upper specification limit
_OBSLSS_ Observed percent of data less than the lower specification limit
_PCHISQ_ p-value for chi-square goodness-of-fit test
_SCALE_ Value of scale parameter for fitted distribution. For the lognormal
distribution, this is the value of specified or estimated according
to the ZETA= option. For all other distributions, this is the value
specified or estimated according to the SIGMA= option.
_SHAPE1_ Value of shape parameter for fitted distribution. For the beta,
gamma, generalized Pareto, and power function distributions, this
is the value of ˛, either specified with the ALPHA= option or
estimated by the procedure. For the lognormal distribution, this is
the value of , either specified with the SIGMA= option or
estimated by the procedure. For the Weibull distribution, this is the
value of c, either specified with the C= option or estimated by the
procedure. For the Johnson SB and SU distributions, this is the
value of ı, either specified with the DELTA= option or estimated
by the procedure. For distributions without a shape parameter
(Gumbel, normal, exponential, and Rayleigh distributions),
_SHAPE1_ is set to missing.
_SHAPE2_ Value of shape parameter for fitted distribution. For the beta
distribution, this is the value of ˇ, either specified with the BETA=
option or estimated by the procedure. For the Johnson SB and SU
distributions, this is the value of , either specified with the
GAMMA= option or estimated by the procedure. For all other
distributions, _SHAPE2_ is set to missing.
_VAR_ Variable name
_WIDTH_ Width of histogram interval
OUTHISTOGRAM= Data Set

The OUTHISTOGRAM= data set contains information about histogram intervals. Because you can specify
multiple HISTOGRAM statements with the CAPABILITY procedure, you can create multiple OUTHIS-
TOGRAM= data sets.
The data set contains a group of observations for each variable plotted with the HISTOGRAM statement.
The group contains an observation for each interval of the histogram, beginning with the leftmost interval
that contains a value of the variable and ending with the rightmost interval that contains a value of the
variable. These intervals will not necessarily coincide with the intervals displayed in the histogram because
the histogram may be padded with empty intervals at either end. If you superimpose one or more fitted
curves on the histogram, the OUTHISTOGRAM= data set contains multiple groups of observations for each
variable (one group for each curve). If you use a BY statement, the OUTHISTOGRAM= data set contains
groups of observations for each BY group. ID variables are not saved in the OUTHISTOGRAM= data set.
The OUTHISTOGRAM= data set contains the variables listed in Table 6.31. By default, an OUTHIS-
TOGRAM= data set contains the _MIDPT_ variable, whose values identify histogram intervals by their
midpoints. When the ENDPOINTS= or NENDPOINTS option is specified, intervals are identified by
endpoint values instead. If the RTINCLUDE option is specified, the _MAXPT_ variable contains an interval’s
upper endpoint value. Otherwise, the _MINPT_ variable contains the interval’s lower endpoint value.
Table 6.31 Variables in the OUTHISTOGRAM= Data Set
_COUNT_ Number of variable values in histogram interval
_CURVE_ Name of fitted distribution (if requested in HISTOGRAM
statement)
_EXPPCT_ Estimated percent of population in histogram interval determined
from optional fitted distribution
_MAXPT_ Upper endpoint of histogram interval
_MIDPT_ Midpoint of histogram interval
_MINPT_ Lower endpoint of histogram interval
_OBSPCT_ Percent of variable values in histogram interval
_VAR_ Variable name
OUTKERNEL= Output Data Set

An OUTKERNEL= data set contains information about kernel density estimates requested with the KERNEL
option. Because you can specify multiple HISTOGRAM statements with the CAPABILITY procedure, you
can create multiple OUTKERNEL= data sets.
An OUTKERNEL= data set contains a group of observations for each kernel density estimate requested
with the HISTOGRAM statement. These observations span a range of analysis variable values recorded
in the _VALUE_ variable. The procedure determines the increment between values, and therefore the
number of observations in the group. The variable _DENSITY_ contains the kernel density calculated for the
corresponding analysis variable value.
When a density curve is overlaid on a histogram, the curve is scaled so that the area under the curve equals the
total area of the histogram bars. The scaled density values are saved in the variable _COUNT_, _PERCENT_,
or _PROPORTION_, depending on the histogram’s vertical axis scale, which is determined by the VSCALE=
option. Only one of these variables appears in a particular OUTKERNEL= data set.
Table 6.32 lists the variables in an OUTKERNEL= data set.
Table 6.32 Variables in the OUTKERNEL= Data Set
_C_ Standardized bandwidth parameter
_COUNT_ Kernel density scaled for VSCALE=COUNT
_DENSITY_ Kernel density
_PERCENT_ Kernel density scaled for VSCALE=PERCENT (default)
_PROPORTION_ Kernel density scaled for VSCALE=PROPORTION
_TYPE_ Kernel function
_VALUE_ Variable value at which kernel function is calculated
_VAR_ Variable name
ODS Tables
Table 6.33 summarizes the ODS tables related to fitted distributions that you can request with the HIS-
TOGRAM statement.
Table 6.33 ODS Tables Produced with the HISTOGRAM

Statement

Bins Histogram bins MIDPERCENTS suboption with
any distribution option, such as
NORMAL( MIDPERCENTS)
FitIndices Capability indices computed INDICES suboption with any
from fitted distribution distribution option, such as
LOGNORMAL( INDICES)
FitQuantiles Quantiles of fitted Any distribution option such as
distribution NORMAL

GoodnessOfFit Goodness-of-fit tests for Any distribution option such as
ParameterEstimates Parameter estimates for fitted Any distribution option such as
distribution NORMAL
Specifications Percents outside Any distribution option such as
specification limits based on NORMAL
empirical and fitted
distributions
ODS Graphics
destination where the graph is produced. HISTOGRAM options used to control the appearance of traditional
When ODS Graphics is in effect, the HISTOGRAM statement assigns a name to the graph it creates. You
can use this name to reference the graph when using ODS. The name is listed in Table 6.34.
Table 6.34 ODS Graphics Produced by the HISTOGRAM Statement

Histogram Histogram
producing charts.
SYMBOL and PATTERN Statement Options

In earlier releases of SAS/QC software, graphical features (such as colors and line types) of specification
lines, histogram bars, and fitted curves were controlled with options in SYMBOL and PATTERN statements
when producing traditional graphics. These options are still supported, although they have been superseded
by options in the HISTOGRAM and SPEC statements. Table 6.35, Table 6.36, and Table 6.37 summarize the
two sets of options. N OTE : These statements have no effect on ODS Graphics output.
Table 6.35 Graphical Enhancement of Histogram Outlines and

Specification Lines
Statement Alternative Statement

Feature and Options and Options
Outline of histogram bars HISTOGRAM statement SYMBOL1 statement
Color CBARLINE=color C=color
Width W=value
Target reference line SPEC statement SYMBOL1 statement
Position TARGET=value
Color CTARGET=color C=color
Line type LTARGET=linetype L=linetype
Width WTARGET=value W=value
Lower specification line SPEC statement SYMBOL2 statement
Position LSL=value
Color CLSL=color C=color
Line type LLSL=linetype L=linetype
Width WLSL=value W=value
Upper specification line SPEC statement SYMBOL3 statement
Position USL=value
Color CUSL=color C=color
Line type LUSL=linetype L=linetype
Width WUSL=value W=value
Table 6.36 Graphical Enhancement of Areas Under Histograms and Curves

Area Under Histogram or Curve and Options and Options
Histogram or curve HISTOGRAM statement PATTERN1 statement
Pattern PFILL=pattern V=pattern
Color CFILL=color C=color
Left of lower specification limit SPEC statement PATTERN2 statement
Pattern PLEFT=pattern V=pattern
Color CLEFT=color C=color
Right of upper specification limit SPEC statement PATTERN3 statement
Pattern PRIGHT=pattern V=pattern
Color CRIGHT=color C=color
Table 6.37 Graphical Enhancement of Fitted Curves

Normal curve Normal-options SYMBOL4 statement
Color COLOR=color C=color
Line type L=linetype L=linetype
Width W=value W=value
Lognormal curve Lognormal-options SYMBOL5 statement
Exponential curve Exponential-options SYMBOL6 statement
Weibull curve Weibull-options SYMBOL7 statement
Gamma curve Gamma-options SYMBOL8 statement
Beta curve Beta-options SYMBOL9 statement
Johnson SB curve SB -options SYMBOL10 statement
Johnson SU curve SU -options SYMBOL11 statement
Rayleigh curve Rayleigh-options SYMBOL12 statement

Generalized Pareto curve Pareto-options SYMBOL13 statement

Gumbel curve Gumbel-options SYMBOL14 statement
Power function curve Power-options SYMBOL15 statement
Inverse Gaussian curve IGauss-options SYMBOL16 statement
Examples: HISTOGRAM Statement

This section provides advanced examples of the HISTOGRAM statement.
Example 6.8: Fitting a Beta Curve

N OTE : See Fitting a Beta Curve on a Histogram in the SAS/QC Sample Library.
You can use a beta distribution to model the distribution of a quantity that is known to vary between lower
and upper bounds. In this example, a manufacturing company uses a robotic arm to attach hinges on metal
sheets. The attachment point should be offset 10.1 mm from the left edge of the sheet. The actual offset
varies between 10.0 and 10.5 mm due to variation in the arm. Offsets for 50 attachment points are saved in
the following data set:
data Measures;
input Length @@;
label Length = 'Attachment Point Offset in mm';
datalines;
10.147 10.070 10.032 10.042 10.102
10.034 10.143 10.278 10.114 10.127
10.122 10.018 10.271 10.293 10.136
10.240 10.205 10.186 10.186 10.080
10.158 10.114 10.018 10.201 10.065
10.061 10.133 10.153 10.201 10.109
10.122 10.139 10.090 10.136 10.066
Examples: HISTOGRAM Statement F 377
10.074 10.175 10.052 10.059 10.077

10.211 10.122 10.031 10.322 10.187
10.094 10.067 10.094 10.051 10.174
;
The following statements create a histogram with a fitted beta density curve:
ods graphics off;

legend2 frame cframe=ligr cborder=black position=center;
title1 'Fitted Beta Distribution of Offsets';
proc capability data=Measures;
specs usl=10.25 lusl=20 cusl=salmon cright=yellow pright=solid;
histogram Length /
beta(theta=10 scale=0.5 color=blue fill)
cfill = ywh
cframe = ligr
href = 10
hreflabel = 'Lower Bound'
lhref = 2
legend = legend2
vaxis = axis1;
axis1 label=(a=90 r=0);
inset n = 'Sample Size'
beta(pchisq = 'P-Value') / pos=ne cfill=ywh;
run;
The histogram is shown in Output 6.8.1. The THETA= beta-option specifies the lower threshold. The
SCALE= beta-option specifies the range between the lower threshold and the upper threshold (in this case,
0.5 mm). Note that in general, the default THETA= and SCALE= values are zero and one, respectively.
Output 6.8.1 Superimposing a Histogram with a Fitted Beta Curve
The FILL beta-option specifies that the area under the curve is to be filled with the CFILL= color. (If FILL
were omitted, the CFILL= color would be used to fill the histogram bars instead.) The CRIGHT= option in
the SPEC statement specifies the color under the curve to the right of the upper specification limit. If the
CRIGHT= option were not specified, the entire area under the curve would be filled with the CFILL= color.
When a lower specification limit is available, you can use the CLEFT= option in the SPEC statement to
specify the color under the curve to the left of this limit.
The HREF= option draws a reference line at the lower bound, and the HREFLABEL= option adds the label
Lower Bound. The option LHREF=2 specifies a dashed line type. The INSET statement adds an inset with
the sample size and the p-value for a chi-square goodness-of-fit test.
In addition to displaying the beta curve, the BETA option summarizes the curve fit, as shown in Output 6.8.2.
The output tabulates the parameters for the curve, the chi-square goodness-of-fit test whose p-value is shown
in Output 6.8.1, the observed and estimated percents above the upper specification limit, and the observed
and estimated quantiles. For instance, based on the beta model, the percent of offsets greater than the upper
specification limit is 6.6%. For computational details, see the section “Formulas for Fitted Curves” on
page 349.
Output 6.8.2 Summary of Fitted Beta Distribution

Fitted Beta Distribution of Offsets

Fitted Beta Distribution for Length (Attachment Point Offset in mm)
Parameters for Beta

Distribution
Threshold Theta 10
Scale Sigma 0.5
Shape Alpha 2.06832
Shape Beta 6.022479
Mean 10.12782
Std Dev 0.072339
Goodness-of-Fit Tests for Beta Distribution

Percent Outside
Specifications for Beta
Distribution
Upper Limit
USL 10.250000
Obs Pct > USL 8.000000
Est Pct > USL 6.618103
Quantiles for Beta

Distribution
Quantile
1.0 10.0180 10.0124
5.0 10.0310 10.0285
10.0 10.0380 10.0416
25.0 10.0670 10.0718
50.0 10.1220 10.1174
75.0 10.1750 10.1735
90.0 10.2255 10.2292
95.0 10.2780 10.2630
99.0 10.3220 10.3237
Example 6.9: Fitting Lognormal, Weibull, and Gamma Curves

N OTE : See Superimposing Fitted Curves on a Histogram in the SAS/QC Sample Library.
To find an appropriate model for a process distribution, you should consider curves from several distribution
families. As shown in this example, you can use the HISTOGRAM statement to fit more than one type of
distribution and display the density curves on the same histogram.
The gap between two plates is measured (in cm) for each of 50 welded assemblies selected at random from
the output of a welding process assumed to be in statistical control. The lower and upper specification limits
for the gap are 0.3 cm and 0.8 cm, respectively. The measurements are saved in a data set named Plates.
data Plates;
label Gap='Plate Gap in cm';
input Gap @@;
datalines;
0.746 0.357 0.376 0.327 0.485 1.741 0.241 0.777 0.768 0.409
0.252 0.512 0.534 1.656 0.742 0.378 0.714 1.121 0.597 0.231
0.541 0.805 0.682 0.418 0.506 0.501 0.247 0.922 0.880 0.344
0.519 1.302 0.275 0.601 0.388 0.450 0.845 0.319 0.486 0.529
1.547 0.690 0.676 0.314 0.736 0.643 0.483 0.352 0.636 1.080
;
The following statements fit three distributions (lognormal, Weibull, and gamma) and display their density
curves on a single histogram:
ods graphics on;

proc capability data=Plates;
var Gap;
histogram /
midpoints=0.2 to 1.8 by 0.2
lognormal
weibull
gamma
nospeclegend;
inset n mean(5.3) std='Std Dev'(5.3) skewness(5.3)
/ pos = ne header = 'Summary Statistics';
run;
The LOGNORMAL, WEIBULL, and GAMMA options superimpose fitted curves on the histogram in
Output 6.9.1. Note that a threshold parameter D 0 is assumed for each curve. In applications where the
threshold is not zero, you can specify with the THETA= option.
Output 6.9.1 Superimposing a Histogram with Fitted Curves
The LOGNORMAL, WEIBULL, and GAMMA options also produce the summaries for the fitted distributions
shown in Output 6.9.2, Output 6.9.3, and Output 6.9.4.
Output 6.9.2 Summary of Fitted Lognormal Distribution

Fitted Lognormal Distribution for Gap (Plate Gap in cm)

Distribution
Threshold Theta 0
Scale Zeta -0.58375
Mean 0.631932
Std Dev 0.336436
Output 6.9.2 continued

Percent Outside Specifications for Lognormal

Distribution
LSL 0.300000 USL 0.800000

Distribution
Quantile
1.0 0.23100 0.17449
5.0 0.24700 0.24526
10.0 0.29450 0.29407
25.0 0.37800 0.39825
50.0 0.53150 0.55780
75.0 0.74600 0.78129
90.0 1.10050 1.05807
95.0 1.54700 1.26862
99.0 1.74100 1.78313
Output 6.9.2 provides four goodness-of-fit tests for the lognormal distribution: the chi-square test and three
tests based on the EDF (Anderson-Darling, Cramér–von Mises, and Kolmogorov-Smirnov). See “Chi-Square
Goodness-of-Fit Test” on page 364 and “EDF Goodness-of-Fit Tests” on page 364 for more information.
The EDF tests are superior to the chi-square test because they are not dependent on the set of midpoints used
for the histogram.
At the ˛ D 0:10 significance level, all four tests support the conclusion that the two-parameter lognormal
distribution with scale parameter O D 0:58, and shape parameter O D 0:50 provides a good model for the
distribution of plate gaps.
Output 6.9.3 Summary of Fitted Weibull Distribution

Fitted Weibull Distribution for Gap (Plate Gap in cm)
Parameters for Weibull

Distribution
Threshold Theta 0
Scale Sigma 0.719208
Shape C 1.961159
Mean 0.637641
Std Dev 0.339248
Goodness-of-Fit Tests for Weibull Distribution

Cramer-von Mises W-Sq 0.1593728 Pr > W-Sq 0.016
Anderson-Darling A-Sq 1.1569354 Pr > A-Sq <0.010
Percent Outside Specifications for Weibull

Distribution
LSL 0.300000 USL 0.800000
Quantiles for Weibull

Distribution
Quantile
1.0 0.23100 0.06889
5.0 0.24700 0.15817
10.0 0.29450 0.22831
25.0 0.37800 0.38102
50.0 0.53150 0.59661
75.0 0.74600 0.84955
90.0 1.10050 1.10040
95.0 1.54700 1.25842
99.0 1.74100 1.56691
Output 6.9.3 provides two EDF goodness-of-fit tests for the Weibull distribution: the Anderson-Darling
and the Cramér–von Mises tests. (See Table 6.29 for a complete list of the EDF tests available in the
HISTOGRAM statement.) The probability values for the chi-square and EDF tests are all less than 0.10,
indicating that the data do not support a Weibull model.
Output 6.9.4 Summary of Fitted Gamma Distribution

Fitted Gamma Distribution for Gap (Plate Gap in cm)
Parameters for Gamma

Distribution
Threshold Theta 0
Scale Sigma 0.155198
Shape Alpha 4.082646
Mean 0.63362
Std Dev 0.313587
Goodness-of-Fit Tests for Gamma Distribution

Anderson-Darling A-Sq 0.5810661 Pr > A-Sq 0.137
Percent Outside Specifications for Gamma

Distribution
LSL 0.300000 USL 0.800000
Quantiles for Gamma

Distribution
Quantile
1.0 0.23100 0.13326
5.0 0.24700 0.21951
10.0 0.29450 0.27938
25.0 0.37800 0.40404
50.0 0.53150 0.58271
75.0 0.74600 0.80804
90.0 1.10050 1.05392
95.0 1.54700 1.22160
99.0 1.74100 1.57939
Output 6.9.4 provides four goodness-of-fit tests for the gamma distribution. The probability value for the
chi-square test is less than 0.10, indicating that the data do not support a gamma model.
Based on this analysis, the fitted lognormal distribution is the best model for the distribution of plate gaps.
You can use this distribution to calculate useful quantities. For instance, you can compute the probability that
the gap of a randomly sampled plate exceeds the upper specification limit, as follows:
PrŒgap > USL D Pr Z > 1 .log.USL / /

D 1 ˆ 1 .log.USL / /

where Z has a standard normal distribution, and ˆ./ is the standard normal cumulative distribution function.
Note that ˆ./ can be computed with the DATA step function PROBNORM. In this example, USL = 0.8 and
PrŒgap > 0:8 D 0:2352. This value is expressed as a percent (Est Pct > USL) in Output 6.9.2.
Example 6.10: Comparing Goodness-of-Fit Tests

N OTE : See Comparing Goodness-of-Fit Tests in the SAS/QC Sample Library.
A weakness of the chi-square goodness-of-fit test is its dependence on the choice of histogram midpoints. An
advantage of the EDF tests is that they give the same results regardless of the midpoints, as illustrated in this
example.
In Example 6.9, the option MIDPOINTS=0.2 TO 1.8 BY 0.2 was used to specify the histogram midpoints for
Gap. The following statements refit the lognormal distribution by using default midpoints (0.3 to 1.8 by 0.3).
data Plates;
input Gap @@;
datalines;
0.746 0.357 0.376 0.327 0.485 1.741 0.241 0.777 0.768 0.409
0.252 0.512 0.534 1.656 0.742 0.378 0.714 1.121 0.597 0.231
0.541 0.805 0.682 0.418 0.506 0.501 0.247 0.922 0.880 0.344
0.519 1.302 0.275 0.601 0.388 0.450 0.845 0.319 0.486 0.529
1.547 0.690 0.676 0.314 0.736 0.643 0.483 0.352 0.636 1.080
;
title1 'Distribution of Plate Gaps';

proc capability data=Plates noprint;
var Gap;
histogram / lognormal
nospeclegend
odstitle = title;
inset n mean(5.3) std='Std Dev'(5.3) skewness(5.3) /
pos = ne
header = 'Summary Statistics';
run;
The histogram is shown in Output 6.10.1.
Output 6.10.1 Lognormal Curve Fit with Default Midpoints
A summary of the lognormal fit is shown in Output 6.10.2. The p-value for the chi-square goodness-of-fit
test is 0.082. Because this value is less than 0.10 (a typical cutoff level), the conclusion is that the lognormal
distribution is not an appropriate model for the data. This is the opposite conclusion drawn from the chi-
square test in Example 6.9, which is based on a different set of midpoints and has a p-value of 0.2756 (see
Output 6.9.2). Moreover, the results of the EDF goodness-of-fit tests are the same because these tests do not
depend on the midpoints. When available, the EDF tests provide more powerful alternatives to the chi-square
test. For a thorough discussion of EDF tests, refer to D’Agostino and Stephens (1986).
Output 6.10.2 Printed Output for the Lognormal Curve

Distribution of Plate Gaps

Fitted Lognormal Distribution for Gap (Plate Gap in cm)

Distribution
Threshold Theta 0
Scale Zeta -0.58375
Mean 0.631932
Std Dev 0.336436

Example 6.11: Computing Capability Indices for Nonnormal Distributions

N OTE : See Nonnormal Distribution Capability Indices in the SAS/QC Sample Library.
Standard capability indices such as Cpk are generally considered meaningful only if the process output
has a normal (or reasonably normal) distribution. In practice, however, many processes have nonnormal
distributions. This example, which is a continuation of Example 6.9 and Example 6.10, shows how you
can use the HISTOGRAM statement to compute generalized capability indices based on fitted nonnormal
distributions.
The following statements produce printed output that is partially listed in Output 6.11.1 and Output 6.11.2:
data Plates;
input Gap @@;
datalines;
0.746 0.357 0.376 0.327 0.485 1.741 0.241 0.777 0.768 0.409
0.252 0.512 0.534 1.656 0.742 0.378 0.714 1.121 0.597 0.231
0.541 0.805 0.682 0.418 0.506 0.501 0.247 0.922 0.880 0.344
0.519 1.302 0.275 0.601 0.388 0.450 0.845 0.319 0.486 0.529
1.547 0.690 0.676 0.314 0.736 0.643 0.483 0.352 0.636 1.080
;
proc capability data=Plates checkindices(alpha=0.05);

specs lsl=0.3 usl= 0.8;
histogram Gap / lognormal(indices) noplot;
run;
The PROC CAPABILITY statement computes the standard capability indices that are shown in Output 6.11.1.
Output 6.11.1 Standard Capability Indices for Variable Gap

Variable: Gap (Plate Gap in cm)

Cp 0.237112 0.190279 0.283853
CPL 0.316422 0.203760 0.426833
CPU 0.157803 0.059572 0.254586
Cpk 0.157803 0.060270 0.255336
Warning: Normality is rejected for alpha = 0.05 using the Shapiro-Wilk test
The CHECKINDICES option in the PROC statement requests a goodness-of-fit test for normality in conjunc-
tion with the indices and displays the warning that normality is rejected at the significance level ˛ D 0:05.
Example 6.9 concluded that the fitted lognormal distribution summarized in Output 6.9.2 is a good model, so
one might consider computing generalized capability indices based on this distribution. These indices are
requested with the INDICES option and are shown in Output 6.11.2. Formulas and recommendations for
these indices are given in “Indices Using Fitted Curves” on page 367.
Output 6.11.2 Fitted Lognormal Distribution Information
Capability
Indices Based
on Lognormal
Distribution
Cp 0.210804
CPL 0.595156
CPU 0.124927
Cpk 0.124927
Example 6.12: Computing Kernel Density Estimates

N OTE : See Superimposing Kernel Density Estimates in the SAS/QC Sample Library.
This example illustrates the use of kernel density estimates to visualize a nonnormal data distribution.
The effective channel length (in microns) is measured for 1225 field effect transistors. The channel lengths
are saved as values of the variable Length in a SAS data set named Channel:
data Channel;
length Lot $ 16;
input Length @@;
select;
when (_n_ <= 425) Lot='Lot 1';
when (_n_ >= 926) Lot='Lot 3';
otherwise Lot='Lot 2';
end;
datalines;
0.91 1.01 0.95 1.13 1.12 0.86 0.96 1.17 1.36 1.10
0.98 1.27 1.13 0.92 1.15 1.26 1.14 0.88 1.03 1.00
0.98 0.94 1.09 0.92 1.10 0.95 1.05 1.05 1.11 1.15
1.11 0.98 0.78 1.09 0.94 1.05 0.89 1.16 0.88 1.19
1.01 1.08 1.19 0.94 0.92 1.27 0.90 0.88 1.38 1.02
... more lines ...
2.13 2.05 1.90 2.07 2.15 1.96 2.15 1.89 2.15 2.04
1.95 1.93 2.22 1.74 1.91
;
When you use kernel density estimates to explore a data distribution, you should try several choices for the
bandwidth parameter c because this determines the smoothness and closeness of the fit. You can specify a
list of C= values with the KERNEL option to request multiple density estimates, as shown in the following
statements:
title "FET Channel Length Analysis";

histogram Length / kernel(c = 0.25 0.50 0.75 1.00)
odstitle = title;
run;
The display, shown in Output 6.12.1, demonstrates the effect of c. In general, larger values of c yield smoother
density estimates, and smaller values yield estimates that more closely fit the data distribution.
Output 6.12.1 Multiple Kernel Density Estimates
Output 6.12.1 reveals strong trimodality in the data, which are explored further in “Creating a One-Way
Comparative Histogram” on page 286.
Example 6.13: Fitting a Three-Parameter Lognormal Curve

N OTE : See Three-Parameter Lognormal Distribution in the SAS/QC Sample Library.
If you request a lognormal fit with the LOGNORMAL option, a two-parameter lognormal distribution is
assumed. This means that the shape parameter and the scale parameter are unknown (unless specified)
and that the threshold is known (it is either specified with the THETA= option or assumed to be zero).
If it is necessary to estimate in addition to and , the distribution is referred to as a three-parameter
lognormal distribution. The equation for this distribution is the same as the equation given in section
“Lognormal Distribution” on page 353 but the method of maximum likelihood must be modified. This
example shows how you can request a three-parameter lognormal distribution.
A manufacturing process (assumed to be in statistical control) produces a plastic laminate whose strength
must exceed a minimum of 25 psi. Samples are tested, and a lognormal distribution is observed for the
strengths. It is important to estimate to determine whether the process is capable of meeting the strength
requirement. The strengths for 49 samples are saved in the following data set:
data Plastic;
label Strength='Strength in psi';
input Strength @@;
datalines;
30.26 31.23 71.96 47.39 33.93 76.15 42.21
81.37 78.48 72.65 61.63 34.90 24.83 68.93
43.27 41.76 57.24 23.80 34.03 33.38 21.87
31.29 32.48 51.54 44.06 42.66 47.98 33.73
25.80 29.95 60.89 55.33 39.44 34.50 73.51
43.41 54.67 99.43 50.76 48.81 31.86 33.88
35.57 60.41 54.92 35.66 59.30 41.96 45.32
;
The following statements use the LOGNORMAL option in the HISTOGRAM statement to display the fitted
three-parameter lognormal curve shown in Output 6.13.1:
ods graphics off;

title 'Three-Parameter Lognormal Fit';
proc capability data=Plastic noprint;
spec lsl=25 cleft=orange clsl=black;
histogram Strength / lognormal(fill color = paoy
theta = est)
cfill = paoy
cframe = ligr
nolegend;
inset lsl='LSL' lslpct / cfill = ywh pos=nw;
inset lognormal / format=6.2 pos=ne cfill = ywh;
run;
Specifying THETA=EST requests a local maximum likelihood estimate (LMLE) for , as described by Cohen
(1951). This estimate is then used to compute estimates for and as described in “Lognormal Distribution”
on page 353. The sample program CAPL3A illustrates a similar computational method implemented as a
SAS/IML program.
N OTE : See Three-Parameter Weibull Distribution in the SAS/QC Sample Library.
Note that you can specify THETA=EST as a Weibull-option to fit a three-parameter Weibull distribution.
Output 6.13.1 Three-Parameter Lognormal Fit
Example 6.14: Annotating a Folded Normal Curve

N OTE : See Cpk for Folded Normal Distribution in the SAS/QC Sample Library.
This example shows how to display a fitted curve that is not supported by the HISTOGRAM statement.
The offset of an attachment point is measured (in mm) for a number of manufactured assemblies, and the
measurements are saved in a data set named Assembly.
data Assembly;
label Offset = 'Offset (in mm)';
input Offset @@;
datalines;
11.11 13.07 11.42 3.92 11.08 5.40 11.22 14.69 6.27 9.76
9.18 5.07 3.51 16.65 14.10 9.69 16.61 5.67 2.89 8.13
9.97 3.28 13.03 13.78 3.13 9.53 4.58 7.94 13.51 11.43
11.98 3.90 7.67 4.32 12.69 6.17 11.48 2.82 20.42 1.01
3.18 6.02 6.63 1.72 2.42 11.32 16.49 1.22 9.13 3.34
1.29 1.70 0.65 2.62 2.04 11.08 18.85 11.94 8.34 2.07
0.31 8.91 13.62 14.94 4.83 16.84 7.09 3.37 0.49 15.19
5.16 4.14 1.92 12.70 1.97 2.10 9.38 3.18 4.18 7.22
15.84 10.85 2.35 1.93 9.19 1.39 11.40 12.20 16.07 9.23
0.05 2.15 1.95 4.39 0.48 10.16 4.81 8.28 5.68 22.81
0.23 0.38 12.71 0.06 10.11 18.38 5.53 9.36 9.32 3.63
12.93 10.39 2.05 15.49 8.12 9.52 7.77 10.70 6.37 1.91
8.60 22.22 1.74 5.84 12.90 13.06 5.08 2.09 6.41 1.40
15.60 2.36 3.97 6.17 0.62 8.56 9.36 10.19 7.16 2.37
12.91 0.95 0.89 3.82 7.86 5.33 12.92 2.64 7.92 14.06
;
The assembly process is in statistical control, and it is decided to fit a folded normal distribution to the offset
measurements. A variable X has a folded normal distribution if X D jY j, where Y is distributed as N.; /.
The fitted density is
.x /2 .x C /2

1
h.x/ D p exp C exp ; x0
2 2 2 2 2
You can use SAS/IML software to compute preliminary estimates of and based on a method of moments
given by Elandt (1961). These estimates are computed by solving equation (19) of Elandt (1961), which is
given by
2
2
p2 e =2 Œ1 2ˆ. /
2
f ./ D DA
1 C 2
where ˆ./ is the standard normal distribution function, and
xN 2
AD 1 P n 2
n i D1 xi
Then the estimates of and are given by

r P
1 n 2
n iD1 xi
O 0 D O 2
1C
O 0 D O O 0
Begin by using the MEANS procedure to compute the first and second moments and using the DATA step to
compute the constant A.
proc means data = Assembly noprint;

var Offset;
output out=Stat mean=m1 var=var n=n min = min;
run;
* Compute constant A from equation (19) of Elandt (1961) ;

data Stat;
keep m2 a min;
set Stat;
a = (m1*m1);
m2 = ((n-1)/n)*var + a;
a = a/m2;
run;
Next, use the SAS/IML subroutine NLPDD to solve equation (19) by minimizing .f . / A/2 , and compute
O 0 and O 0 .
proc iml;
use Stat;
read all var {m2} into m2;
read all var {a} into a;
read all var {min} into min;
* f(t) is the function in equation (19) of Elandt (1961) ;

start f(t) global(a);
y = .39894*exp(-0.5*t*t);
y = (2*y-(t*(1-2*probnorm(t))))**2/(1+t*t);
y = (y-a)**2;
return(y);
finish;
* Minimize (f(t)-A)**2 and estimate mu and sigma ;

if ( min < 0 ) then do;
print "Warning: Observations are not all nonnegative.";
print " The folded normal is inappropriate.";
stop;
end;
if ( a < 0.637 ) then do;
print "Warning: the folded normal may be inappropriate";
end;
opt = { 0 0 };
con = { 1e-6 };
x0 = { 1.7 };
tc = { . . . . . 1e-12 . . . . . . .};
call nlpdd(rc,etheta0,"f",x0,opt,con,tc);
esig0 = sqrt(m2/(1+etheta0*etheta0));
emu0 = etheta0*esig0;
create Prelim var {emu0 esig0 etheta0};

append;
close Prelim;
* Define the log likelihood of the folded normal ;
start g(p) global(x);
y = 0.0;
do i = 1 to nrow(x);
z = exp( (-0.5/p[2])*(x[i]-p[1])*(x[i]-p[1]) );
z = z + exp( (-0.5/p[2])*(x[i]+p[1])*(x[i]+p[1]) );
y = y + log(z);
end;
y = y - nrow(x)*log( sqrt( p[2] ) );
return(y);
finish;
* Maximize the log likelihood with subroutine NLPDD ;
use Assembly;
read all var {Offset} into x;
esig0sq = esig0*esig0;
x0 = emu0||esig0sq;
opt = { 1 0 };
con = { . 0.0, . . };
call nlpdd(rc,xr,"g",x0,opt,con);
emu = xr[1];
esig = sqrt(xr[2]);
etheta = emu/esig;
create Parmest var{emu esig etheta};
append;
close Parmest;
quit;
title 'The Data Set Prelim';

proc print data=Prelim noobs;
run;
The preliminary estimates are saved in the data set Prelim, as shown in Output 6.14.1.
Output 6.14.1 Preliminary Estimates of , , and

The Data Set Prelim
EMU0 ESIG0 ETHETA0

6.51735 6.54953 0.99509
Now, using O 0 and O 0 as initial estimates, call the NLPDD subroutine to maximize the log likelihood, l.; /,
of the folded normal distribution, where, up to a constant,
n
/2 .xi C /2

X .xi
l.; / D n log C log exp C exp
2 2 2 2
i D1
* Define the log likelihood of the folded normal ;

start g(p) global(x);
y = 0.0;
do i = 1 to nrow(x);
z = exp( (-0.5/p[2])*(x[i]-p[1])*(x[i]-p[1]) );
z = z + exp( (-0.5/p[2])*(x[i]+p[1])*(x[i]+p[1]) );
y = y + log(z);
end;
y = y - nrow(x)*log( sqrt( p[2] ) );
return(y);
finish;
* Maximize the log likelihood with subroutine NLPDD ;

use assembly;
read all var {offset} into x;
esig0sq = esig0*esig0;
x0 = emu0||esig0sq;
opt = { 1 0 };
con = { . 0.0, . . };
call nlpdd(rc,xr,"g",x0,opt,con);
emu = xr[1];
esig = sqrt(xr[2]);
etheta = emu/esig;
create parmest var{emu esig etheta};

append;
close parmest;
quit;
title 'The Data Set PARMEST';

proc print data=Parmest noobs;
var emu esig etheta;
run;
The data set Parmest saves the maximum likelihood estimates O and O (as well as =
O ),
O as shown in
Output 6.14.2.
Output 6.14.2 Final Estimates of , , and

The Data Set PARMEST
EMU ESIG ETHETA

6.66761 6.39650 1.04239
To annotate the curve on a histogram, begin by computing the width and endpoints of the histogram intervals.
The following statements save these values in an OUTFIT= data set called OUT. Note that a plot is not
produced at this point.
ods graphics off;

proc capability data = Assembly noprint;
histogram Offset / outfit = Out normal(noprint) noplot;
run;
title 'OUTFIT= Data Set Out';

proc print data=Out noobs round;
var _var_ _curve_ _locatn_ _scale_ _chisq_ _df_ _pchisq_
_midpt1_ _width_ _midptn_ _expect_ _eststd_ _adasq_
_adp_ _cvmwsq_ _cvmp_ _ksd_ _ksp_;
run;
Output 6.14.3 provides a partial listing of the data set Out. The width and endpoints of the histogram bars are
saved as values of the variables _WIDTH_, _MIDPT1_, and _MIDPTN_. See “Output Data Sets” on page 369.
Output 6.14.3 The OUTFIT= Data Set Out

OUTFIT= Data Set Out
_VAR_ _CURVE_ _LOCATN_ _SCALE_ _CHISQ_ _DF_ _PCHISQ_ _MIDPT1_ _WIDTH_ _MIDPTN_
Offset NORMAL 7.62 5.24 31.17 5 0 1.5 3 22.5
_EXPECT_ _ESTSTD_ _ADASQ_ _ADP_ _CVMWSQ_ _CVMP_ _KSD_ _KSP_

7.62 5.24 1.9 0.01 0.28 0.01 0.09 0.01
The following statements create an annotate data set named Anno, which contains the coordinates of the
fitted curve:
data Anno;
merge Parmest Out;
length function color $ 8;
function = 'point';
color = 'black';
size = 2;
xsys = '2';
ysys = '2';
when = 'a';
constant = 39.894*_width_;;
left = _midpt1_ - .5*_width_;
right = _midptn_ + .5*_width_;
inc = (right-left)/100;
do x = left to right by inc;
z1 = (x-emu)/esig;
z2 = (x+emu)/esig;
y = (constant/esig)*(exp(-0.5*z1*z1)+exp(-0.5*z2*z2));
output;
function = 'draw';
end;
run;
The following statements read the ANNOTATE= data set and display the histogram and fitted curve, as
shown in Output 6.14.4:
ods graphics off;

title "Folded Normal Distribution";
proc capability data=Assembly noprint;
spec usl=27 cusl=black lusl=2 wusl=2;
histogram Offset / annotate = Anno;
run;
Output 6.14.4 Histogram with Annotated Folded Normal Curve
INSET Statement: CAPABILITY Procedure
Overview: INSET Statement

Graphical displays such as histograms and probability plots are commonly used for process capability
analysis. You can use the INSET statement to enhance these plots by adding a box or table (referred to as
an inset) of summary statistics directly to the graph. An inset typically displays statistics calculated by the
CAPABILITY procedure but can also display values provided in a SAS data set. A typical application of
the INSET statement is to augment a histogram with the sample size, mean, standard deviation, and process
capability index Cpk .
Note that the INSET statement by itself does not produce a display and must be used with the CDFPLOT,
COMPHISTOGRAM, HISTOGRAM, PPPLOT, PROBPLOT, or QQPLOT statement.
You can use options in the INSET statement to
specify the position of the inset

Getting Started: INSET Statement F 399
specify a header for the inset table
specify graphical enhancements, such as background colors, text colors, text height, text font, and drop
shadows
The INSET statement is not applicable when you produce line printer plots by specifying the LINEPRINTER
option in the PROC CAPABILITY statement.
Getting Started: INSET Statement

This section introduces the INSET statement with examples that illustrate commonly used options. Complete
syntax for the INSET statement is presented in the section “Syntax: INSET Statement” on page 403, and
advanced examples are given in the section “Examples: INSET Statement” on page 423.
Displaying Summary Statistics on a Histogram

N OTE : See Histograms with INSET Statement Features in the SAS/QC Sample Library.
In a plant producing copper wire, an important quality characteristic is the torsion strength, measured as the
twisting force in pounds per inch necessary to break the wire. The following statements create the SAS data
set Wire, which contains the torsion strengths (Strength) for 50 different wire samples:
data Wire;
label Strength='Torsion Strength in lb/in';
input Strength @@;
datalines;
25 25 36 31 26 36 29 37 37 20
34 27 21 35 30 41 33 21 26 26
19 25 14 32 30 29 31 26 22 24
34 33 28 26 43 30 40 32 32 31
25 26 27 34 33 27 33 29 30 31
;
A histogram is used to examine the data distribution. For a more complete report, the sample size, minimum
value, maximum value, mean, and standard deviation are displayed on the histogram. The following
statements illustrate how to inset these statistics:
ods graphics off;

title 'Torsion Strength of Copper Wire';
proc capability data=Wire noprint;
spec lsl=22 usl=38;
histogram Strength;
inset n min max mean std;
run;
The resulting histogram is displayed in Figure 6.17. The INSET statement immediately follows the plot
statement that creates the graphical display (in this case, the HISTOGRAM statement). Specify the keywords
for inset statistics (such as N, MIN, MAX, MEAN, and STD) immediately after the word INSET. The inset
statistics appear in the order in which you specify the keywords.
Figure 6.17 A Histogram with an Inset
A complete list of keywords that you can use with the INSET statement is provided in “Summary of INSET
Keywords” on page 405. Note that the set of keywords available for a particular display depends on both the
plot statement that precedes the INSET statement and the options that you specify in the plot statement.
The following examples illustrate options commonly used for enhancing the appearance of an inset.
Formatting Values and Customizing Labels

By default, each inset statistic is identified with an appropriate label, and each numeric value is printed using
an appropriate format. However, you may want to provide your own labels and formats. For example, in
Figure 6.17 the default format for the standard deviation prints an excessive number of decimal places. The
following statements correct this problem, as well as customizing some of the labels displayed in the inset:
ods graphics on;

spec lsl=22 usl=38;
histogram Strength;
inset n='Sample Size' min max mean std='Std Dev' (5.2);
run;
Getting Started: INSET Statement F 401
Graphics, so the histogram is created using ODS Graphics instead of traditional graphics.
The resulting histogram is displayed in Figure 6.18. You can provide your own label by specifying the
keyword for that statistic followed by an equal sign (=) and the label in quotes. The label can have up to 24
characters.
The format 5.2 specified in parentheses after the keyword STD displays the standard deviation with a field
width of five and two decimal places. In general, you can specify any numeric SAS format in parentheses
after an inset keyword. You can also specify a format to be used for all the statistics in the INSET statement
with the FORMAT= option (see the next section, “Adding a Header and Positioning the Inset” on page 402).
For more information about SAS formats, refer to SAS Formats and Informats: Reference.
Note that if you specify both a label and a format for a statistic, the label must appear before the format, as
with the keyword STD in the previous statements.
Figure 6.18 Formatting Values and Customizing Labels in an Inset

Adding a Header and Positioning the Inset

In the previous examples, the inset is displayed in the upper left corner of the plot, the default position for
insets added to histograms. You can control the inset position with the POSITION= option. In addition, you
can display a header at the top of the inset with the HEADER= option. The following statements create the
chart shown in Figure 6.19:
ods graphics off;

spec lsl=22 usl=38;
histogram Strength;
inset n='Sample Size' min max range mode sum mean
std='Standard Deviation' var stdmean skewness
kurtosis / format = 6.1
pos = rm
header = 'Data Summary' cfill = ywh;
run;
The header (in this case, Data Summary) can be up to 40 characters. Note that a long list of inset statistics is
requested. Consequently, POSITION=RM is specified to position the inset in the right margin. For more
information about positioning, see “Details: INSET Statement” on page 418. Also note that the FORMAT=
option is used to format all inset statistics. The options, such as HEADER=, POSITION=, and FORMAT=,
are specified after the slash (/) in the INSET statement. For more details on INSET statement options, see
Syntax: INSET Statement F 403
Figure 6.19 Adding a Header and Repositioning the Inset
Syntax: INSET Statement

The syntax for the INSET statement is as follows:
INSET keyword-list < / options > ;
You can use any number of INSET statements in the CAPABILITY procedure. Each INSET statement pro-
duces an inset and must follow one of the plot statements: CDFPLOT, COMPHISTOGRAM, HISTOGRAM,
PPPLOT, PROBPLOT, or QQPLOT. The inset appears in all displays produced by the plot statement that
immediately precedes it. The statistics are displayed in the order in which they are specified. For example,
the following statements produce a cumulative distribution plot with two insets and a histogram with one
inset:
proc capability data=Wire;

cdfplot Strength;
inset mean std min max n;
inset p1 p5 p10;
histogram Strength;
inset var skewness kurtosis;
run;
The statistics displayed in an inset are computed for a specific process variable from observations for the
current BY group. For example, in the following statements, there are two process variables (Strength and
Diameter) and a BY variable (Batch). If there are three different batches (levels of Batch), then a total of six
histograms are produced. The statistics in each inset are computed for a particular variable and batch. The
labels in the inset are the same for each histogram.
proc capability data=Wire2;

by Batch;
histogram Strength Diameter / normal;
inset mean std min max normal(mu sigma);
run;
The components of the INSET statement are described as follows.
keyword-list
can include any of the keywords listed in “Summary of INSET Keywords” on page 405. Some
keywords allow secondary keywords to be specified in parentheses immediately after the primary
keyword. Also, some inset statistics are available only if you request plot statements and options for
which those statistics are calculated. For example, consider the following statements:

histogram Strength / normal;
inset mean std normal(ad adpval);
run;
The keywords MEAN and STD display the sample mean and standard deviation of Strength. The
primary keyword NORMAL with the secondary keywords AD and ADPVAL display the Anderson-
Darling goodness-of-fit test statistic and p-value in the inset as well. The statistics specified with the
NORMAL keyword are available only because a normal distribution has been fit to the data by using
the NORMAL option in the HISTOGRAM statement. See the section “Summary of INSET Keywords”
for a list of available keywords.
Typically, you specify keywords, to display statistics computed by the CAPABILITY procedure.
However, you can also specify the keyword DATA= followed by the name of a SAS data set to display
customized statistics. This data set must contain two variables:
a character variable named _LABEL_ whose values provide labels for inset entries.
a variable named _VALUE_, which can be either character or numeric, and whose values provide
values for inset entries.
The label and value from each observation in the DATA= data set occupy one line in the inset. The
position of the DATA= keyword in the keyword list determines the position of its lines in the inset.
By default, inset statistics are identified with appropriate labels, and numeric values are printed using
appropriate formats. However, you can provide customized labels and formats. You provide the
customized label by specifying the keyword for that statistic followed by an equal sign (=) and the label
in quotes. Labels can have up to 24 characters. You provide the numeric format in parentheses after
the keyword. Note that if you specify both a label and a format for a statistic, the label must appear
before the format. For an example, see “Formatting Values and Customizing Labels” on page 400.
options
appear after the slash (/) and control the appearance of the inset. For example, the following INSET
statement uses two appearance options (POSITION= and CTEXT=):
inset mean std min max / position=ne ctext=yellow;
The POSITION= option determines the location of the inset, and the CTEXT= option specifies the
color of the text of the inset.
See “Summary of Options” on page 414 for a list of all available options, and “Dictionary of Options”
on page 415 for detailed descriptions. Note the difference between keywords and options; keywords
specify the information to be displayed in an inset, whereas options control the appearance of the inset.
Summary of INSET Keywords

Summary Statistics and Process Capability Indices
Table 6.38 Summary Statistics
Keyword Description
CSS Corrected sum of squares
CV Coefficient of variation
GEOMEAN Geometric mean
HARMEAN Harmonic mean
KURTOSIS | KURT Kurtosis
MAX Largest value
MEAN Sample mean
MIN Smallest value
MODE Most frequent value
N Sample size
NEXCL Number of observations excluded by MAXNBIN= or
MAXSIGMAS= option
NMISS Number of missing values
NOBS Number of observations
RANGE Range
SKEWNESS | SKEW Skewness
STD | STDDEV Standard deviation
STDMEAN | STDERR Standard error of the mean
SUM Sum of the observations
SUMWGT Sum of the weights
USS Uncorrected sum of squares
VAR Variance
Table 6.39 Percentile Statistics
Keyword Description
P1 1st percentile
P5 5th percentile
P10 10th percentile
Q1 | P25 Lower quartile (25th percentile)
MEDIAN | Q2 | P50 Median (50th percentile)
Q3 | P75 Upper quartile (75th percentile)
P90 90th percentile
P95 95th percentile
P99 99th percentile
QRANGE Interquartile range (Q3–Q1)
Table 6.40 lists keywords for distribution-free confidence limits for percentiles requested with the CIPCTLDF
option.
Table 6.40 Keywords for Distribution-Free Confidence

Limits for Percentiles
Keyword Description
P1_LCL_DF 1st percentile lower confidence limit
P1_UCL_DF 1st percentile upper confidence limit
P5_LCL_DF 5th percentile lower confidence limit
P5_UCL_DF 5th percentile upper confidence limit
Q1_LCL_DF | P25_LCL_DF Lower quartile (25th percentile) lower confidence
limit
Q1_UCL_DF | P25_UCL_DF Lower quartile (25th percentile) upper confidence
limit
MEDIAN_LCL_DF | Q2_LCL_DF | Median (50th percentile) lower confidence limit
P50_LCL_DF
MEDIAN_UCL_DF | Q2_UCL_DF | Median (50th percentile) upper confidence limit
P50_UCL_DF
Q3_LCL_DF | P75_LCL_DF Upper quartile (75th percentile) lower confidence limit
Q3_UCL_DF | P75_UCL_DF Upper quartile (75th percentile) upper confidence limit
Table 6.41 lists keywords for percentile confidence limits computed assuming normality requested with the
CIPCTLNORMAL option.
Table 6.41 Keywords for Percentile Confidence Limits

Assuming Normality
Keyword Description
P1_LCL 1st percentile lower confidence limit
P1_UCL 1st percentile upper confidence limit
P5_LCL 5th percentile lower confidence limit
P5_UCL 5th percentile upper confidence limit
Q1_LCL | P25_LCL Lower quartile (25th percentile) lower confidence limit
Q1_UCL | P25_UCL Lower quartile (25th percentile) upper confidence limit
MEDIAN_LCL | Q2_LCL | Median (50th percentile) lower confidence limit
P50_LCL
MEDIAN_UCL | Q2_UCL | Median (50th percentile) upper confidence limit
P50_UCL
Q3_LCL | P75_LCL Upper quartile (75th percentile) lower confidence limit
Q3_UCL | P75_UCL Upper quartile (75th percentile) upper confidence limit
Table 6.42 Robust Statistics
Keyword Description
GINI Gini’s mean difference
MAD Median absolute difference about the median
QN Qn , alternative to MAD
SN Sn , alternative to MAD
STD_GINI Gini’s standard deviation
STD_MAD MAD standard deviation
STD_QN Qn standard deviation
STD_QRANGE Interquartile range standard deviation
STD_SN Sn standard deviation
Table 6.43 Hypothesis Testing
Keyword Description
MSIGN Sign statistic
NORMALTEST Test statistic for normality
PNORMAL Probability value for the test of normality
SIGNRANK Signed rank statistic
PROBM Probability of greater absolute value for the sign statistic
PROBN Probability value for the test of normality
PROBS Probability value for the signed rank test
PROBT Probability value for the Student’s t test
T Statistics for Student’s t test
Table 6.44 Input Data Set
Keyword Description
DATA= (label, value) pairs from input data set
Table 6.45 Capability Indices and Confidence Limits
Keyword Description
CP Capability index Cp
CPLCL Lower confidence limit for Cp
CPUCL Upper confidence limit for Cp
CPK Capability index Cpk
CPKLCL Lower confidence limit for Cpk
CPKUCL Upper confidence limit for Cpk
CPL Capability index CPL
CPM Capability index Cpm
CPMLCL Lower confidence limit for Cpm
CPMUCL Upper confidence interval for Cpm
CPU Capability index CPU
K Capability index K
Table 6.46 Specification Limits and Related Information
Keyword Description
LSL Lower specification limit
USL Upper specification limit
TARGET Target value
PCTGTR Percent of nonmissing observations that exceed the
upper specification limit
Keyword Description
PCTLSS Percent of nonmissing observations that are less than
the lower specification limit
PCTBET Percent of nonmissing observations between the upper
and lower specification limits (inclusive)
Statistics Available with Parametric Density Estimates

You can request parametric density estimates with all plot statements in the CAPABILITY procedure
(CDFPLOT, COMPHISTOGRAM, HISTOGRAM, PPPLOT, PROBPLOT, and QQPLOT). You can display
parameters and statistics associated with these estimates in an inset by specifying a distribution keyword
followed by secondary keywords in parentheses. For example, the following statements create a histogram
for Strength with a fitted exponential density curve:

histogram Strength / exp;
inset exp(sigma theta);
run;
The secondary keywords SIGMA and THETA for the EXP distribution keyword request an inset displaying
the values of the exponential scale parameter and threshold parameter . You must request the distribution
option in the plot statement to display the corresponding distribution statistics in an inset. Specifying a
distribution keyword with no secondary keywords produces an inset displaying the full set of parameters for
that distribution. See Output 6.15.1 for an example of an inset with statistics from a fitted normal curve.
The following table describes the available distribution keywords. Note that some keywords are not available
with all plot statements.
Table 6.47 Density Estimation Primary Keywords
Keyword Distribution Plot Statement Availability

BETA Beta All but COMPHISTOGRAM
EXPONENTIAL Exponential All but COMPHISTOGRAM
GAMMA Gamma All but COMPHISTOGRAM
GUMBEL Gumbel All but COMPHISTOGRAM
IGAUSS Inverse Gaussian CDFPLOT, HISTOGRAM, PPPLOT
LOGNORMAL Lognormal All but COMPHISTOGRAM
NORMAL Normal All
PARETO Generalized Pareto All but COMPHISTOGRAM
POWER Power function All but COMPHISTOGRAM
RAYLEIGH Rayleigh All but COMPHISTOGRAM
SB Johnson SB HISTOGRAM
SU Johnson SU HISTOGRAM
WEIBULL Weibull All but COMPHISTOGRAM
WEIBULL2 2-parameter Weibull PROBPLOT, QQPLOT
Table 6.48 lists the secondary keywords available with each distribution keyword listed in Table 6.47. In
many cases, aliases can be used (for example, ALPHA in place of SHAPE1).
Table 6.48 Density Estimation Secondary Keywords
Secondary
Keyword Alias Description
Secondary Keywords Available with the BETA Keyword
ALPHA SHAPE1 First shape parameter ˛
BETA SHAPE2 Second shape parameter ˇ
SIGMA SCALE Scale parameter
THETA THRESHOLD Lower threshold parameter
MEAN Mean of the fitted distribution
STD Standard deviation of the fitted distribution
Secondary Keywords Available with the EXPONENTIAL Keyword

THETA THRESHOLD Threshold parameter
Secondary Keywords Available with the GAMMA Keyword

ALPHA SHAPE Shape parameter ˛
Secondary Keywords Available with the GUMBEL Keyword

MU Location parameter
Secondary Keywords Available with the IGAUSS Keyword

MU Mean parameter
LAMBDA Shape parameter
Secondary Keywords Available with the LOGNORMAL Keyword

SIGMA SHAPE Shape parameter
ZETA SCALE Scale parameter
Secondary
Secondary Keywords Available with the NORMAL Keyword
MU MEAN Mean parameter
SIGMA STD Scale parameter
Secondary Keywords Available with the PARETO Keyword

ALPHA Shape parameter ˛
Secondary Keywords Available with the POWER Keyword

ALPHA Shape parameter ˛
Secondary Keywords Available with the RAYLEIGH Keyword

Secondary Keywords Available with the SB Keyword

DELTA SHAPE1 Shape parameter ı
GAMMA SHAPE2 Shape parameter
Secondary Keywords Available with the SU Keyword

DELTA SHAPE1 Shape parameter ı
GAMMA SHAPE2 Shape parameter
THETA Location parameter
Secondary Keywords Available with the WEIBULL Keyword

C SHAPE Shape parameter c
Secondary
Secondary Keywords Available with the WEIBULL2 Keyword

C SHAPE Shape parameter c
THETA THRESHOLD Known lower threshold 0
The secondary keywords listed in Table 6.49 can be used with any distribution keyword but only with the
HISTOGRAM and COMPHISTOGRAM plot statements.
Table 6.49 Statistics Computed from Any Parametric

Density Estimate
Secondary
Keyword Description
CPK Capability index Cpk
ESTPCTLSS Estimated percentage less than the lower specification limit
ESTPCTGTR Estimated percentage greater than the upper specification limit
K Capability index K
The secondary keywords listed in Table 6.50 can be used with any distribution keyword but only with the
HISTOGRAM plot statement (see Example 6.15).
Table 6.50 Goodness-of-Fit Statistics for Fitted Curves
Secondary
Keyword Description
CHISQ Chi-square statistic
DF Degrees of freedom for the chi-square test
PCHISQ Probability value for the chi-square test
AD Anderson-Darling EDF test statistic
ADPVAL Anderson-Darling EDF test p-value
CVM Cramér–von Mises EDF test statistic
Secondary
Keyword Description
CVMPVAL Cramér–von Mises EDF test p-value
KSD Kolmogorov-Smirnov EDF test statistic
KSDPVAL Kolmogorov-Smirnov EDF test p-value
Table 6.51 lists primary keywords available only with the HISTOGRAM and COMPHISTOGRAM plot
statements. These keywords display fill areas on a histogram. If you fit a parametric density on a histogram
and request that the area under the curve be filled, these keywords display the percentage of the distribution
area that lies below the lower specification limit, between the specification limits, or above the upper
specification limit. If you do not fill the area beneath a parametric density estimate, these keywords display
the observed proportion of observations (that is, the area in the bars of the histogram).
You should use these options with the FILL, CFILL=, and PFILL= options in the HISTOGRAM and
COMPHISTOGRAM statements and with the CLEFT=, CRIGHT=, PLEFT=, and PRIGHT= options in the
SPEC statements. See Output 6.16.1 for an example.
Table 6.51 Curve Area Keywords

BETWEENPCT BETPCT Area between the specification limits
LSLPCT Area below the lower specification limit
USLPCT Area above the upper specification limit
Statistics Available with Nonparametric Kernel Density Estimates

You can request nonparametric kernel density estimates with the HISTOGRAM and COMPHISTOGRAM
plot statements. You can display statistics associated with these estimates by specifying a kernel density
keyword followed by secondary keywords in parentheses. For example, the following statements create a
histogram for Strength with a fitted kernel density estimate:

histogram Strength / kernel;
inset kernel(c amise);
run;
The secondary keywords C and AMISE for the KERNEL keyword display the values of the standardized
bandwidth c and the approximate mean integrated square error.
Note that you can specify more than one kernel density estimate on a single histogram. If you specify
multiple kernel density estimates, you can request inset statistics for all of the estimates with the KERNEL
keyword, or you can display inset statistics for up to five individual curves with KERNELn keywords, as in
the following example:

histogram Strength / kernel(c = 1 2 3);
inset kernel2(c) kernel3(c);
run;
Three kernel density estimates are displayed on the histogram, but the inset displays the value of c only for
the second and third estimates.
Table 6.52 lists the kernel density keywords. Table 6.53 lists the available secondary keywords.
Table 6.52 Kernel Density Estimate Primary Keywords
Keyword Description
KERNEL Displays statistics for all kernel estimates
KERNELn Displays statistics for only the nth kernel density estimate
n D 1; 2; 3; 4; or 5
Table 6.53 Secondary Keywords Available with the KERNEL Keyword
Secondary Keyword Description

TYPE Kernel type: normal, quadratic, or triangular
BANDWIDTH Bandwidth for the density estimate
BWIDTH Alias for BANDWIDTH
C Standardized bandwidth c for the density estimate:
15
cDQ n where n = sample size, D bandwidth, and
Q = interquartile range
AMISE Approximate mean integrated square error (MISE) for the
kernel density
Summary of Options
The following table lists the INSET statement options. See the section “Dictionary of Options” for complete
descriptions of the options.
Table 6.54 INSET Options
Option Description
CFILL= Specifies color of inset background
CFILLH= Specifies color of header background
CFRAME= Specifies color of frame
CHEADER= Specifies color of header text
CSHADOW= Specifies color of drop shadow
CTEXT= Specifies color of inset text
DATA Specifies data units for POSITION=.x; y/
coordinates
Option Description
FONT= Specifies font of text
FORMAT= Specifies format of values in inset
GUTTER= Specifies gutter width for inset in top or
bottom margin
HEADER= Specifies header text
HEIGHT= Specifies height of inset text
NCOLS= Specifies number of columns for inset in top
or bottom margin
NOFRAME Suppresses frame around inset
POSITION= Specifies position of inset
REFPOINT= Specifies reference point of inset positioned
with POSITION=.x; y/ coordinates
The following sections provide detailed descriptions of options for the INSET statement. Terms used in this
section are illustrated in Figure 6.20.
Figure 6.20 The Inset
General Options
You can specify the following general options:
DATA
specifies that data coordinates are to be used in positioning the inset with the POSITION= option.
The DATA option is available only when you specify POSITIOND .x; y/, and it must be placed
immediately after the coordinates .x; y/. For details, see the entry for the POSITION= option or
“Positioning the Inset Using Coordinates” on page 420. See Figure 6.23 for an example.
FORMAT=format
specifies a format for all the values displayed in an inset. If you specify a format for a particular statistic,
then this format overrides the format you specified with the FORMAT= option. See Figure 6.19 or
Output 6.15.1 for an example.
GUTTER=value
specifies the gutter width in percent screen units for an inset located in the top or bottom margin of
ODS Graphics output. The gutter is the space between columns of (label, value) pairs in an inset. The
default value is four. This option is ignored if ODS Graphics is disabled.
HEADER= ‘string’
specifies the header text. The string cannot exceed 40 characters. If you do not specify the HEADER=
option, no header line appears in the inset. If all the keywords listed in the INSET statement are
secondary keywords corresponding to a fitted curve on a histogram, a default header is displayed that
indicates the distribution and identifies the curve. See Figure 6.19 for an example of a specified header
and Output 6.15.1 for an example of the default header for a fitted normal curve.
NCOLS=n
specifies the number of columns of (label, value) pairs displayed in an inset located in the top or bottom
margin of ODS Graphics output. The default value is three. This option is ignored if ODS Graphics is
disabled.
NOFRAME
suppresses the frame drawn around the text.
POSITION=position
POS=position
determines the position of the inset. The position can be a compass point keyword, a margin keyword,
or a pair of coordinates .x; y/. You can specify coordinates in axis percent units or axis data units. For
more information, see “Details: INSET Statement” on page 418. By default, POSITION=NW, which
positions the inset in the upper left (northwest) corner of the display.
N OTE : In this release of the CAPABILITY procedure, you cannot specify coordinates with the
POSITION= option when producing ODS Graphics output.

CFILL=color | BLANK
specifies the color of the background (including the header background if you do not specify the
CFILLH= option). See Output 6.15.1 for an example.
If you do not specify the CFILL= option, then by default, the background is empty. This means that
items that overlap the inset (such as curves, histogram bars, or specification limits) show through the
inset. If you specify any value for the CFILL= option, then overlapping items no longer show through
the inset. Specify CFILL=BLANK to leave the background uncolored and also to prevent items from
showing through the inset.
CFILLH=color
specifies the color of the header background. By default, if you do not specify a CFILLH= color, the
CFILL= color is used.
CFRAME=color
specifies the color of the frame. By default, the frame is the same color as the axis of the plot.
CHEADER=color
specifies the color of the header text. By default, if you do not specify a CHEADER= color, the
CTEXT= color is used.
CSHADOW=color
CS=color
specifies the color of the drop shadow. See Output 6.16.1 for an example. By default, if you do not
specify the CSHADOW= option, a drop shadow is not displayed.
CTEXT=color
CT=color
specifies the color of the text. By default, the inset text color is the same as the other text on the plot.
FONT=font
specifies the font of the text. By default, the font is SIMPLEX if the inset is located in the interior of
the plot, and the font is the same as the other text displayed on the plot if the inset is located in the
exterior of the plot.
HEIGHT=value
specifies the height of the text.
REFPOINT=BR | BL | TR | TL
RP=BR | BL | TR | TL
specifies the reference point for an inset that is positioned by a pair of coordinates with the POSITION=
option. Use the REFPOINT= option with POSITION= coordinates. The REFPOINT= option specifies
which corner of the inset frame you want positioned at coordinates .x; y/. The keywords BL, BR, TL,
and TR represent bottom left, bottom right, top left, and top right, respectively. See Figure 6.24 for an
example. The default is REFPOINT=BL.
If you specify the position of the inset as a compass point or margin keyword, the REFPOINT= option
is ignored. For more information, see “Positioning the Inset Using Coordinates” on page 420.
Details: INSET Statement

This section provides details on three different methods of positioning the inset with the POSITION= option.
With the POSITION= option, you can specify
compass points
keywords for margin positions
coordinates in data units or percent axis units
Positioning the Inset Using Compass Points

N OTE : See Positioning the Inset in the SAS/QC Sample Library.
You can specify the eight compass points N, NE, E, SE, S, SW, W, and NW as keywords for the POSITION=
option. The following statements create the display in Figure 6.21, which demonstrates all eight compass
positions. The default is NW.
ods graphics off;

histogram Strength / odstitle = title;
inset n / header='Position = NW' pos=nw;
inset mean / header='Position = N ' pos=n ;
inset sum / header='Position = NE' pos=ne;
inset max / header='Position = E ' pos=e ;
inset min / header='Position = SE' pos=se;
inset nobs / header='Position = S ' pos=s ;
inset range / header='Position = SW' pos=sw;
inset mode / header='Position = W ' pos=w ;
run;
Details: INSET Statement F 419
Figure 6.21 Insets Positioned Using Compass Points
Positioning the Inset in the Margins

You can also position the inset in one of the four margins surrounding the plot area using the margin keywords
LM, RM, TM, or BM, as illustrated in Figure 6.22.
Figure 6.22 Positioning Insets in the Margins
For an example of an inset placed in the right margin, see Figure 6.19. Margin positions are recommended if
a large number of statistics are listed in the INSET statement. If you attempt to display a lengthy inset in the
interior of the plot, it is likely that the inset will collide with the data display.
Positioning the Inset Using Coordinates

If you are producing traditional graphics, you can also specify the position of the inset with coordinates:
POSITIOND .x; y/. The coordinates can be given in axis percent units (the default) or in axis data units.
N OTE : In this release of the CAPABILITY procedure, you cannot position insets by using coordinates when
producing ODS Graphics output.
Data Unit Coordinates

If you specify the DATA option immediately following the coordinates, the inset is positioned using axis data
units. For example, the following statements place the bottom left corner of the inset at 12.5 on the horizontal
axis and 10 on the vertical axis:
Details: INSET Statement F 421

histogram Strength;
inset n / header = 'Position=(12.5,10)'
position = (12.5,10) data;
run;
The histogram is displayed in Figure 6.23. By default, the specified coordinates determine the position of the
bottom left corner of the inset. You can change this reference point with the REFPOINT= option, as in the
next example.
Figure 6.23 Inset Positioned Using Data Unit Coordinates
Axis Percent Unit Coordinates

If you do not use the DATA option, the inset is positioned using axis percent units. The coordinates of the
bottom left corner of the display are .0; 0/, while the upper right corner is .100; 100/. For example, the
following statements create a histogram with two insets, both positioned using coordinates in axis percent
units:

histogram Strength;
inset min / position = (5,25)
header = 'Position=(5,25)'
refpoint = tl;
inset max / position = (95,95)
header = 'Position=(95,95)'
refpoint = tr;
run;
The display is shown in Figure 6.24. Notice that the REFPOINT= option is used to determine which corner
of the inset is to be placed at the coordinates specified with the POSITION= option. The first inset has
REFPOINT=TL, so the top left corner of the inset is positioned 5% of the way across the horizontal axis and
25% of the way up the vertical axis. The second inset has REFPOINT=TR, so the top right corner of the
inset is positioned 95% of the way across the horizontal axis and 95% of the way up the vertical axis. Note
also that coordinates in axis percent units must be between 0 and 100.
Figure 6.24 Inset Positioned Using Axis Percent Unit Coordinates

Examples: INSET Statement F 423
Examples: INSET Statement

This section provides advanced examples that use the INSET statement.
Example 6.15: Inset for Goodness-of-Fit Statistics

N OTE : See Inset for Goodness-of-Fit Statistics in the SAS/QC Sample Library.
This example fits a normal curve to the torsion strength data used in the section “Getting Started: INSET
Statement” on page 399. The following statements fit a normal curve and request an inset summarizing the
fitted curve with the mean, the standard deviation, and the Anderson-Darling goodness-of-fit test:

spec lsl=22 usl=38;
histogram Strength / normal(noprint)
nocurvelegend
odstitle = title;
inset normal(mu sigma ad adpval) / format = 7.2;
run;
The resulting histogram is displayed in Output 6.15.1. The NOCURVELEGEND option in the HISTOGRAM
statement suppresses the default legend for curve parameters.
Output 6.15.1 Inset Table with Normal Curve Information
Example 6.16: Inset for Areas Under a Fitted Curve

N OTE : See Inset for Areas Under a Fitted Curve in the SAS/QC Sample Library.
You can use the INSET keywords LSLPCT, USLPCT, and BETWEENPCT to inset legends for areas under
histogram bars or fitted curves. The following statements create a histogram with an inset legend for the
shaded area under the fitted normal curve to the left of the lower specification limit:
ods graphics off;

legend2 FRAME CFRAME=ligr CBORDER=black POSITION=center;
spec lsl=22 llsl=2 clsl=black cleft=vibg
usl=38 lusl=20 cusl=black;
histogram Strength / cframe = ligr
cfill = bibg
legend = legend2
normal(color=black noprint fill);
inset lsl='LSL' lslpct / cfill=ywh cshadow=dagr;
run;
Examples: INSET Statement F 425
The histogram is displayed in Output 6.16.1. The LSLPCT keyword in the INSET statement requests a
legend for the area under the curve to the left of the lower specification limit. The CLEFT= option is used to
fill the area under the normal curve to the left of the line, and the CFILL= color is used to fill the remaining
area. If the FILL normal-option were not specified, the CLEFT= and CFILL= colors would be applied to the
corresponding areas under the histogram, not the normal curve, and the inset box would reflect the area under
the histogram bars.
You can use the USLPCT keyword in the INSET statement to request a legend for the area to the right of
an upper specification limit, and you can use the BETWEENPCT keyword to request a legend for the area
between the lower and upper limits. By default, the legend requested with each of the keywords LSLPCT,
USLPCT, and BETWEENPCT displays a rectangle that matches the color of the corresponding area. You
can substitute a customized label for each rectangle by specifying the keyword followed by an equal sign (=)
and the label in quotes.
Output 6.16.1 Displaying Areas Under the Normal Curve

INTERVALS Statement: CAPABILITY Procedure
Overview: INTERVALS Statement

The INTERVALS statement tabulates various statistical intervals for selected process variables. The types of
intervals you can request include the following:
intervals that are computed assuming the data are sampled from a normal population—see Hahn and
Meeker (1991) for a detailed discussion of these intervals:
– approximate simultaneous prediction intervals for future observations

– prediction intervals for the mean of future observations
– statistical tolerance intervals that contain at least a specified proportion of the population
– confidence intervals for the population mean
– prediction intervals for the standard deviation of future observations
– confidence intervals for the population standard deviation
intervals that are computed without any assumption about the distribution of the population—see
Krishnamoorthy and Mathew (2009) for a detailed discussion of these intervals.
– nonparametric statistical tolerance intervals that contain at least a specified proportion of the
population
You can use options in the INTERVALS statement to
specify which intervals to compute
provide probability or confidence levels for intervals
suppress printing of output tables
create an output data set containing interval information
specify interval type (one-sided lower, one-sided upper, or two-sided)

Getting Started: INTERVALS Statement F 427
Getting Started: INTERVALS Statement

This section introduces the INTERVALS statement with simple examples that illustrate commonly used
options. Complete syntax for the INTERVALS statement is presented in the section “Syntax: INTERVALS
Statement” on page 429.
Computing Statistical Intervals

N OTE : See Calculating Various Statistical Intervals in the SAS/QC Sample Library.
The following statements create the data set Cans, which contains measurements (in ounces) of the fluid
weights of 100 drink cans. The filling process is assumed to be in statistical control.
data Cans;
label Weight = "Fluid Weight (ounces)";
input Weight @@;
datalines;
12.07 12.02 12.00 12.01 11.98 11.96 12.04 12.05 12.01 11.97
12.03 12.03 12.00 12.04 11.96 12.02 12.06 12.00 12.02 11.91
12.05 11.98 11.91 12.01 12.06 12.02 12.05 11.90 12.07 11.98
12.02 12.11 12.00 11.99 11.95 11.98 12.05 12.00 12.10 12.04
12.06 12.04 11.99 12.06 11.99 12.07 11.96 11.97 12.00 11.97
12.09 11.99 11.95 11.99 11.99 11.96 11.94 12.03 12.09 12.03
11.99 12.00 12.05 12.04 12.05 12.01 11.97 11.93 12.00 11.97
12.13 12.07 12.00 11.96 11.99 11.97 12.05 11.94 11.99 12.02
11.95 11.99 11.91 12.06 12.03 12.06 12.05 12.04 12.03 11.98
12.05 12.05 12.11 11.96 12.00 11.96 11.96 12.00 12.01 11.98
;
Note that this data set is introduced in “Computing Descriptive Statistics” on page 207 of “PROC CAPABIL-
ITY and General Statements” on page 205. The analysis in that section provides evidence that the weight
measurements are normally distributed.
By default, the INTERVALS statement computes and prints the six intervals described in the entry for the
METHODS= option. The following statements tabulate these intervals for the variable Weight:
title 'Statistical Intervals for Fluid Weight';

proc capability data=Cans noprint;
intervals Weight;
run;
The intervals are displayed in Figure 6.26.
Figure 6.25 Statistical Intervals for Weight

Statistical Intervals for Fluid Weight

Two-Sided Statistical Intervals for Weight Assuming Normality
Approximate Prediction
Interval Containing All of k
Future Observations
Prediction
Confidence k Limits
99.00% 1 11.89 12.13
99.00% 2 11.87 12.14
99.00% 3 11.87 12.15
95.00% 1 11.92 12.10
95.00% 2 11.90 12.12
95.00% 3 11.89 12.12
90.00% 1 11.93 12.09
90.00% 2 11.92 12.10
90.00% 3 11.91 12.11
Prediction Interval
Containing the Mean of k
Future Observations
Prediction
Confidence k Limits
99.00% 1 11.89 12.13
99.00% 2 11.92 12.10
99.00% 3 11.94 12.08
95.00% 1 11.92 12.10
95.00% 2 11.94 12.08
95.00% 3 11.95 12.06
90.00% 1 11.93 12.09
90.00% 2 11.95 12.06
90.00% 3 11.96 12.05
Tolerance Interval Containing

At Least Proportion p of the
Population
Tolerance
Confidence p Limits
99.00% 0.900 11.92 12.10
99.00% 0.950 11.90 12.12
99.00% 0.990 11.86 12.15
95.00% 0.900 11.92 12.10
95.00% 0.950 11.90 12.11
95.00% 0.990 11.87 12.15
90.00% 0.900 11.92 12.09
90.00% 0.950 11.91 12.11
90.00% 0.990 11.88 12.14
Syntax: INTERVALS Statement F 429
Confidence Limits
Containing the Mean
Confidence
Confidence Limits
99.00% 11.997 12.022
95.00% 12.000 12.019
90.00% 12.002 12.017
Prediction Interval Containing

the Standard Deviation of k
Future Observations
Prediction
Confidence k Limits
99.00% 2 0.0003 0.1348
99.00% 3 0.0033 0.1110
95.00% 2 0.0015 0.1069
95.00% 3 0.0075 0.0919
90.00% 2 0.0030 0.0932
90.00% 3 0.0106 0.0825
Confidence Limits
Containing the Standard
Deviation
Confidence
Confidence Limits
99.00% 0.040 0.057
95.00% 0.041 0.055
90.00% 0.042 0.053
Syntax: INTERVALS Statement

The syntax for the INTERVALS statement is as follows:
INTERVALS < variables > < / options > ;
You can specify INTERVAL as an alias for INTERVALS. You can use any number of INTERVALS statements
in the CAPABILITY procedure. The components of the INTERVALS statement are described as follows.
variables
gives a list of variables for which to compute intervals. If you specify a VAR statement, the variables
must also be listed in the VAR statement. Otherwise, the variables can be any numeric variable in the
input data set. If you do not specify a list of variables, then by default the INTERVALS statement
computes intervals for all variables in the VAR statement (or all numeric variables in the input data set
if you do not use a VAR statement).
options
alter the defaults for computing and printing intervals and for creating output data sets.
Summary of Options
The following tables list the INTERVALS statement options by function. For complete descriptions, see
Table 6.55 INTERVAL Statement Options
Option Description
ALPHA= Specifies probability or confidence levels associated with the
intervals
K= Specifies values of k for prediction intervals
METHODS= Specifies which intervals are computed
NOPRINT Suppresses the output tables
OUTINTERVALS= Specifies an output data set containing interval information
P= Specifies values of p for tolerance intervals
TYPE= Specifies the type of intervals (one-sided lower, one-sided upper, or
two-sided)
The following entries provide detailed descriptions of options in the INTERVALS statement.
ALPHA=value-list
specifies values of ˛, the probability or confidence associated with the interval. For example, the
following statements tabulate the default intervals at probability or confidence levels of ˛ D 0:05,
˛ D 0:10, ˛ D 0:15, and ˛ D 0:20:

intervals width / alpha = 0.05 0.10 0.15 0.20;
run;
Note that some references use D 1 ˛ to denote probability or confidence levels. Values for the
ALPHA= option must be between 0.00001 to 0.99999. By default, values of 0.01, 0.05, and 0.10 are
used.
K=value-list
specifies values of k for prediction intervals. Default values of 1, 2, and 3 are used for the prediction
interval for k future observations and for the prediction interval for the mean of k future observations.
Default values of 2 and 3 are used for the prediction interval for the standard deviation of k future
observations. The values must be integers.
Syntax: INTERVALS Statement F 431
METHODS=indices
METHOD=indices
specifies which intervals are computed. The indices can range from 1 to 7, and they correspond to the
intervals described in Table 6.56.
Table 6.56 Intervals Computed for METHOD=Index
Option Description
1 Approximate simultaneous prediction interval for k future observations
2 Prediction interval for the mean of k future observations
3 Statistical tolerance interval that contains at least proportion p of the
population
4 Confidence interval for the population mean
5 Prediction interval for the standard deviation of k future observations
6 Confidence interval for the population standard deviation
7 Nonparametric statistical tolerance interval that contains at least proportion
p of the population
For example, the following statements tabulate confidence limits for the population mean
(METHOD=4) and confidence limits for the population standard deviation (METHOD=6):

intervals width / methods=4 6;
run;
Formulas for the intervals are given in “Methods for Computing Statistical Intervals” on page 432. By
default, the procedure computes the first six intervals, for which it assumes that the data are sampled
from a normal population.
NOPRINT
suppresses the tables produced by default. This option is useful when you only want to save the interval
information in an OUTINTERVALS= data set.
OUTINTERVALS=SAS-data-set
OUTINTERVAL=SAS-data-set
OUTINT=SAS-data-set
specifies an output SAS data set containing the intervals and related information. For example, the
following statements create a data set named ints containing intervals for the variable width:

intervals width / outintervals=ints;
run;
See “OUTINTERVALS= Data Set” on page 435 for details.

P=value-list
specifies values of p for the tolerance intervals. These values must be between 0.00001 to 0.99999.
Note that the P= option applies only to the tolerance intervals (a value of 3 or 7, or both, for the
METHODS= option). By default, values of 0.90, 0.95, and 0.99 are used.
TYPE=LOWER | UPPER | TWOSIDED

determines whether the intervals computed are one-sided lower, one-sided upper, or two-sided intervals,
respectively. See Example 6.17 for an example. By default, TYPE=TWOSIDED.
Details: INTERVALS Statement

formulas for statistical intervals
OUTINTERVALS= data sets
Methods for Computing Statistical Intervals

The formulas for statistical intervals that are described in this section use the notation shown in Table 6.57.
Table 6.57 Notation Used in This Section
Notation Definition
n Number of nonmissing values for a variable
XN Mean of variable
s Standard deviation of variable
z˛ 100˛th percentile of the standard normal distribution
t˛ ./ 100˛th percentile of the central t distribution with degrees of freedom
t˛0 .ı; / 100˛th percentile of the noncentral t distribution with noncentrality
parameter ı and degrees of freedom
F˛ .1 ; 2 / 100˛th percentile of the F distribution with 1 degrees of freedom in
the numerator and 2 degrees of freedom in the denominator
2˛ ./ 100˛th percentile of the 2 distribution with degrees of freedom
2˛ .ı; / 100˛th percentile of the noncentral 2 distribution with noncentrality
parameter ı and degrees of freedom
X.i / ith order statistic from a sample X D .X1 ; X2 ; : : : ; Xn /
The values of the variable are assumed to be independent and normally distributed. The intervals are
computed using the degrees of freedom as the divisor for the standard deviation s. This divisor corresponds
to the default of VARDEF=DF in the PROC CAPABILITY statement. If you specify another value for the
VARDEF= option, intervals are not computed.
You select the intervals to be computed with the METHODS= option. The next six sections give computational
details for each of the METHODS= options.
Details: INTERVALS Statement F 433
METHODS=1
This requests an approximate simultaneous prediction interval for k future observations. Two-sided intervals
are computed using the conservative approximations
q
1
Lower Limit D XN t1 ˛
2k
.n 1/s 1 C n
q
1
Upper Limit D XN C t1 ˛
2k
.n 1/s 1 C n
One-sided limits are computed using the conservative approximation
q
1
k
.n 1/s 1 C n
q
1
k
.n 1/s 1 C n
Hahn (1970b) states that these approximations are satisfactory except for combinations of small n, large k,
and large ˛. Refer also to Hahn (1969, 1970a) and Hahn and Meeker (1991).
METHODS=2
This requests a prediction interval for the mean of k future observations. Two-sided intervals are computed as
q
1 1
2
.n 1/s k
C n
q
1 1
2
.n 1/s k
C n
One-sided limits are computed as
q
1 1
Lower Limit D XN t1 ˛ .n 1/s k
C n
q
1 1
Upper Limit D XN C t1 ˛ .n 1/s k
C n
METHODS=3
This requests a statistical tolerance interval that contains at least proportion p of the population. Two-sided
intervals are computed as
Lower Limit D XN ks
Upper Limit D XN C ks
where k is the solution of the integral equation
!
1 1/2p .z 2 ; 1/
r
2n .n
Z
1 2
P 2n 1 > e 2 nz dz D 1 ˛
0 k2
Lower Limit D XN g 0 .p I n I 1 ˛/s
Upper Limit D XN C g 0 .p I n I 1 ˛/s

p
where g 0 .p I n I 1 ˛/ D p1 t 0
n 1 ˛ .zp n; n 1/.
For a thorough discussion of tolerance intervals and tolerance limits, see Krishnamoorthy and Mathew (2009).
METHODS=4
This requests a confidence interval for the population mean. Two-sided intervals are computed as
2
.n 1/ psn
2
.n 1/ psn
Lower Limit D XN t1 ˛ .n 1/ psn
Upper Limit D XN C t1 ˛ .n 1/ psn
METHODS=5
This requests a prediction interval for the standard deviation of k future observations. Two-sided intervals are
computed as
1
2
Lower Limit D s F1 ˛
2
.n 1; k 1/
12
Upper Limit D s F1 ˛
2
.k 1; n 1/
1
Lower Limit D s .F1 ˛ .n 1; k 1// 2
1
Upper Limit D s .F1 ˛ .k 1; n 1// 2
METHODS=6
This requests a confidence interval for the population standard deviation. Two-sided intervals are computed
as
r
n 1
Lower Limit D s
21 ˛ .n 1/
2
r
n 1
Upper Limit D s
2˛ .n 1/
2
Details: INTERVALS Statement F 435
q
n 1
Lower Limit D s
21 ˛ .n 1/
q
n 1
Upper Limit D s
2˛ .n 1/
METHODS=7
This requests a nonparametric statistical tolerance interval that contains at least proportion p of the population.
Nonparametric tolerance intervals are based on order statistics, as described by Krishnamoorthy and Mathew
(2009).
Two-sided intervals are computed as
Lower Limit D X.r/ ; r D d n 2 k e
Upper Limit D X.s/ ; s D n r C1

where Y is a random variable from a binomial.n; p/ distribution and k is the least integer for which
P .Y k 1/ 1 ˛.
Lower Limit D X.k/
Upper Limit D X.n kC1/
where Y is a random variable from a binomial.n; 1 p/ distribution and k is the largest integer for which
P .Y k/ 1 ˛.
OUTINTERVALS= Data Set

Each INTERVALS statement can create an output data set specified with the OUTINTERVALS= option. The
OUTINTERVALS= data set contains statistical intervals and related parameters.
The number of observations in the OUTINTERVALS= data set depends on the number of variables analyzed,
the number of tests specified, and the results of the tests. The OUTINTERVALS= data set is constructed as
follows:
The OUTINTERVALS= data set contains a group of observations for each variable analyzed.
Each group contains one or more observations for each interval you specify with the METHODS=
option. The actual number depends upon the number of combinations of the ALPHA=, K=, and P=
values.
The following variables are saved in the OUTINTERVALS= data set:

_ALPHA_ Value of ˛ associated with the intervals
_K_ Value of K= for the prediction intervals
_LOWER_ Lower endpoint of interval
_METHOD_ Interval index (1–7)
_P_ Value of P= for the tolerance intervals
_TYPE_ Type of interval (ONESIDED or TWOSIDED)
_UPPER_ Upper endpoint of interval
_VAR_ Variable name
If you use a BY statement, the BY variables are also saved in the OUTINTERVALS= data set.
ODS Tables
Table 6.58 summarizes the ODS tables that you can request with the INTERVALS statement.
Table 6.58 ODS Tables Produced with the INTERVALS Statement

Intervals1 Prediction interval for future observations METHODS=1
Intervals2 Prediction interval for mean METHODS=2
Intervals3 Tolerance interval for proportion of population METHODS=3
Intervals4 Confidence limits for mean METHODS=4
Intervals5 Prediction interval for standard deviation METHODS=5
Intervals6 Confidence limits for standard deviation METHODS=6
Intervals7 Nonparametric tolerance interval for proportion of population METHODS=7
Examples: INTERVALS Statement

Example 6.17: Computing One-Sided Lower Prediction Limits
N OTE : See Calculating Various Statistical Intervals in the SAS/QC Sample Library.
This example uses the Cans data set from the section “Getting Started: INTERVALS Statement” on page 427.
You can specify options after the slash (/) in the INTERVALS statement to control the computation and
printing of intervals. The following statements produce a table of one-sided lower prediction limits for the
mean, which is displayed in Output 6.17.1:

intervals Weight / methods = 1 2
type = lower;
run;
The METHODS= option specifies which intervals to compute, and the TYPE= option requests one-sided
lower limits. All the options available in the INTERVALS statement are listed in “Summary of Options” on
page 430 and are described in “Dictionary of Options” on page 430.
Examples: INTERVALS Statement F 437
Output 6.17.1 One-Sided Lower Prediction Limits for the Mean


One-Sided Lower Statistical Intervals for Weight Assuming Normality
Approximate Prediction
Limit For All of k Future
Observations
Lower
Confidence k Limit
99.00% 1 11.90
99.00% 2 11.89
99.00% 3 11.88
95.00% 1 11.93
95.00% 2 11.92
95.00% 3 11.91
90.00% 1 11.95
90.00% 2 11.93
90.00% 3 11.92
Prediction Limit For the

Mean of k Future
Observations
Lower
Confidence k Limit
99.00% 1 11.90
99.00% 2 11.93
99.00% 3 11.94
95.00% 1 11.93
95.00% 2 11.95
95.00% 3 11.96
90.00% 1 11.95
90.00% 2 11.97
90.00% 3 11.97
Example 6.18: Computing Nonparametric Tolerance Limits

The following statements produce tables of tolerance limits that assume normality and nonparametric
tolerance limits; the tables are shown in Output 6.18.1.

intervals Weight / methods = 3 7;
run;
Output 6.18.1 Tolerance Limits


Two-Sided Statistical Intervals for Weight Assuming Normality

Population
Tolerance
Confidence p Limits
99.00% 0.900 11.92 12.10
99.00% 0.950 11.90 12.12
99.00% 0.990 11.86 12.15
95.00% 0.900 11.92 12.10
95.00% 0.950 11.90 12.11
95.00% 0.990 11.87 12.15
90.00% 0.900 11.92 12.09
90.00% 0.950 11.91 12.11
90.00% 0.990 11.88 12.14

Two-Sided Nonparametric Statistical Intervals for Weight

Population
Tolerance
Confidence p Limits
99.00% 0.900 11.91 12.11
99.00% 0.950 . .
99.00% 0.990 . .
95.00% 0.900 11.91 12.11
95.00% 0.950 11.90 12.13
95.00% 0.990 . .
90.00% 0.900 11.91 12.11
90.00% 0.950 11.90 12.13
90.00% 0.990 . .
OUTPUT Statement: CAPABILITY Procedure F 439
OUTPUT Statement: CAPABILITY Procedure
Overview: OUTPUT Statement

You can use the OUTPUT statement to save summary statistics in a SAS data set. This information can then
be used to create customized reports or to save historical information about a process.
You can use options in the OUTPUT statement to
specify the statistics to save in the output data set
specify the name of the output data set
compute and save percentiles not automatically computed by the CAPABILITY procedure
Getting Started: OUTPUT Statement

This section introduces the OUTPUT statement with simple examples that illustrate commonly used options.
Complete syntax for the OUTPUT statement is presented in the section “Syntax: OUTPUT Statement” on
page 442, and advanced examples are given in the section “Examples: OUTPUT Statement” on page 449.
Saving Summary Statistics in an Output Data Set

N OTE : See Saving CAPABILITY Output in a Data Set in the SAS/QC Sample Library.
An automobile manufacturer producing seat belts saves summary information in an output data set with the
CAPABILITY procedure. The following statements create the data set Belts, which contains the breaking
strengths (Strength) and widths (Width) of a sample of 50 belts:
data Belts;
label Strength = 'Breaking Strength (lb/in)'
Width = 'Width in Inches';
input Strength Width @@;
datalines;
1243.51 3.036 1221.95 2.995 1131.67 2.983 1129.70 3.019
1198.08 3.106 1273.31 2.947 1250.24 3.018 1225.47 2.980
1126.78 2.965 1174.62 3.033 1250.79 2.941 1216.75 3.037
1285.30 2.893 1214.14 3.035 1270.24 2.957 1249.55 2.958
1166.02 3.067 1278.85 3.037 1280.74 2.984 1201.96 3.002
1101.73 2.961 1165.79 3.075 1186.19 3.058 1124.46 2.929
1213.62 2.984 1213.93 3.029 1289.59 2.956 1208.27 3.029
1247.48 3.027 1284.34 3.073 1209.09 3.004 1146.78 3.061
1224.03 2.915 1200.43 2.974 1183.42 3.033 1195.66 2.995
1258.31 2.958 1136.05 3.022 1177.44 3.090 1246.13 3.022
1183.67 3.045 1206.50 3.024 1195.69 3.005 1223.49 2.971
1147.47 2.944 1171.76 3.005 1207.28 3.065 1131.33 2.984
1215.92 3.003 1202.17 3.058
;
The following statements produce two output data sets containing summary statistics:
proc capability data=Belts;

var Strength Width;
output out=Means mean=smean wmean;
output out=Strstats mean=smean std=sstd min=smin max=smax;
run;
proc print data=Means;

run;
proc print data=Strstats;

run;
Note that if you specify an OUTPUT statement, you must also specify a VAR statement. You can use multiple
OUTPUT statements with a single procedure statement. Each OUTPUT statement creates a new data set.
The OUT= option specifies the name of the output data set. In this case, two data sets, Means and Strstats,
are created. See Figure 6.26 for a listing of Means and Figure 6.27 for a listing of Strstats.
Summary statistics are saved in an output data set by specifying keyword=names after the OUT= option. In
the preceding statements, the first OUTPUT statement specifies the keyword MEAN followed by the names
smean and wmean. The second OUTPUT statement specifies the keywords MEAN, STD, MIN, and MAX,
for which the names smean, sstd, smin, and smax are given.
The keyword specifies the statistic to be saved in the output data set, and the names determine the names for
the new variables. The first name listed after a keyword contains that statistic for the first variable listed in
the VAR statement; the second name contains that statistic for the second variable in the VAR statement, and
so on.
Thus, the data set Means contains the mean of Strength in a variable named smean and the mean of Width in
a variable named wmean. The data set Strstats contains the mean, standard deviation, minimum value, and
maximum value of Strength in the variables smean, sstd, smin, and smax, respectively.
Figure 6.26 Listing of the Output Data Set Means
Obs smean wmean

1 1205.75 3.00584
Figure 6.27 Listing of the Output Data Set Strstats
Obs smean sstd smax smin

1 1205.75 48.3290 1289.59 1101.73
Getting Started: OUTPUT Statement F 441
Saving Percentiles in an Output Data Set

N OTE : See Saving CAPABILITY Output in a Data Set in the SAS/QC Sample Library.
The CAPABILITY procedure automatically computes the 1st, 5th, 10th, 25th, 75th, 90th, 95th, and 99th
percentiles for each variable. You can save these percentiles in an output data set by specifying the appropriate
keywords. For example, the following statements create an output data set named Pctlstr containing the 5th
and 95th percentiles of the variable Strength:
proc capability data=Belts noprint;

var Strength Width;
output out=Pctlstr p5=p5str p95=p95str;
run;
proc print data=Pctlstr;

run;
The output data set Pctlstr is listed in Figure 6.28.
Figure 6.28 Listing of the Output Data Set Pctlstr
Obs p95str p5str

1 1284.34 1126.78
You can use the PCTLPTS=, PCTLPRE=, and PCTLNAME= options to save percentiles not automatically
computed by the CAPABILITY procedure. For example, the following statements create an output data set
named Pctls containing the 20th and 40th percentiles of the variables Strength and Width:
proc capability data=Belts noprint;

var Strength Width;
output out=Pctls pctlpts = 20 40
pctlpre = S W
pctlname = pct20 pct40;
run;
proc print data=Pctls;

run;
The PCTLPTS= option specifies the percentiles to compute (in this case, the 20th and 40th percentiles). The
PCTLPRE= and PCTLNAME= options build the names for the variables containing the percentiles. The
PCTLPRE= option gives prefixes for the new variables, and the PCTLNAME= option gives a suffix to add to
the prefix. Note that if you use the PCTLPTS= specification, you must also use the PCTLPRE= specification.
For details on these options, see the section “Syntax: OUTPUT Statement” on page 442.
The preceding OUTPUT statement saves the 20th and 40th percentiles of Strength and Width in the variables
spct20, wpct20, spct40, and wpct40. The output data set Pctls is listed in Figure 6.29.
Figure 6.29 Listing of the Output Data Set Pctls
Obs Spct20 Wpct20 Spct40 Wpct40

1 1165.91 2.9595 1199.26 2.995
Syntax: OUTPUT Statement

The syntax for the OUTPUT statement is as follows:
OUTPUT < OUT=SAS-data-set > < keyword1=names . . . keywordk=names > < percentile-options > ;
You can use any number of OUTPUT statements in the CAPABILITY procedure. Each OUTPUT statement
creates a new data set containing the statistics specified in that statement. When you use the OUTPUT
statement, you must also use the VAR statement. In addition, the OUTPUT statement must contain at least
one of the following:
a specification of the form keyword=names
the PCTLPTS= and PCTLPRE= options
You can use the OUT= option to specify the name of the output data set:
OUT=SAS-data-set
specifies the name of the output data set. To create a permanent SAS data set, specify a two-level name.
See SAS DATA Step Statements: Reference for more information on permanent SAS data sets. For
example, the previous statements create an output data set named Summary. If the OUT= option is
omitted, then by default the new data set is named using the DATAn convention.
A keyword=names specification selects a statistic to be included in the output data set and gives names to the
new variables that contain the statistics. Specify a keyword for each desired statistic, an equal sign, and the
names of the variables to contain the statistic.
In the output data set, the first variable listed after a keyword in the OUTPUT statement contains the statistic
for the first variable listed in the VAR statement; the second variable contains the statistic for the second
variable in the VAR statement, and so on. The list of names following the equal sign can be shorter than the
list of variables in the VAR statement. In this case, the procedure uses the names in the order in which the
variables are listed in the VAR statement. Consider the following example:

var length width height;
output out=summary mean=mlength mwidth;
run;
The variables mlength and mwidth contain the means for length and width. The mean for height is computed
by the procedure but is not saved in the output data set.
Table 6.59 lists all keywords available in the OUTPUT statement grouped by type. Formulas for selected
statistics are given in the section “Details: CAPABILITY Procedure” on page 230.
Table 6.59 OUTPUT Statement Statistic Keywords
Keyword Description
Descriptive Statistics
CSS Sum of squares corrected for the mean
CV Percent coefficient of variation
Syntax: OUTPUT Statement F 443
Keyword Description
GEOMEAN Geometric mean
HARMEAN Harmonic mean
KURTOSIS | KURT Kurtosis
MAX Largest (maximum) value
MEAN Mean
MIN Smallest (minimum) value
MODE Most frequent value (if not unique, the smallest mode)
N Number of observations on which calculations are based
NMISS Number of missing values
NOBS Number of observations
RANGE Range
SKEWNESS | SKEW Skewness
STD | STDDEV Standard deviation
STDMEAN | STDERR Standard error of the mean
SUM Sum
SUMWGT Sum of weights
USS Uncorrected sum of squares
VAR Variance
Quantile Statistics
MEDIAN | P50 | Q2 Median (50th percentile)
P1 1st percentile
P5 5th percentile
P10 10th percentile
P90 90th percentile
P95 95th percentile
P99 99th percentile
Q1 | P25 Lower quartile (25th percentile)
Q3 | P75 Upper quartile (75th percentile)
QRANGE Interquartile range (Q3–Q1)
Robust Statistics
GINI Gini’s mean difference
MAD Median absolute difference
QN 2nd variation of median absolute difference
SN 1st variation of median absolute difference
STD_GINI Standard deviation for Gini’s mean difference
STD_MAD Standard deviation for median absolute difference
STD_QN Standard deviation for the second variation of the median absolute
difference
STD_QRANGE Estimate of the standard deviation, based on interquartile range
STD_SN Standard deviation for the first variation of the median absolute
difference
Keyword Description
Hypothesis Test Statistics
MSIGN Sign statistic
NORMAL Test statistic for normality. If the sample size is less than or equal to
2000, this is the Shapiro-Wilk W statistic. Otherwise, it is the
Kolmogorov D statistic.
PNORMAL | PROBN p-value for normality test
PROBM Probability of a greater absolute value for the sign statistic
PROBS Probability of a greater absolute value for the signed rank statistic
PROBT Two-tailed p-value for Student’s t statistic with n 1 degrees of
freedom
SIGNRANK Signed rank statistic
T Student’s t statistic to test the null hypothesis that the population mean
is equal to 0
Specification Limits and Related Statistics

LSL Lower specification limit
PCTGTR Percent of nonmissing observations greater than
the upper specification limit
PCTLSS Percent of nonmissing observations less than
the lower specification limit
TARGET Target value
USL Upper specification limit
Capability Indices and Related Statistics

CPLCL Lower confidence limit for Cp
CPUCL Upper confidence limit for Cp
CPK Capability index Cpk (also denoted CPK)
CPKLCL Lower confidence limit for Cpk
CPKUCL Upper confidence limit for Cpk
CPLLCL Lower confidence limit for CPL
CPLUCL Upper confidence limit for CPL
CPMLCL Lower confidence limit for Cpm
CPMUCL Upper confidence limit for Cpm
CPULCL Lower confidence limit for CPU
CPUUCL Upper confidence limit for CPU
K Capability index k (also denoted K)
The CAPABILITY procedure automatically computes the 1st, 5th, 10th, 25th, 50th, 75th, 90th, 95th, and
99th percentiles for the data. You can save these statistics in an output data set by using keyword=names
specifications. You can request additional percentiles by using the PCTLPTS= option. The following
percentile-options are related to these additional percentiles:
CIPCTLDF=(cipctl-options)
CIQUANTDF=(cipctl-options)
requests distribution-free confidence limits for percentiles that are requested with the PCTLPTS=
option. In other words, no specific parametric distribution such as the normal is assumed for the data.
PROC CAPABILITY uses order statistics (ranks) to compute the confidence limits as described by
Hahn and Meeker (1991). This option does not apply if you use a WEIGHT statement. You can specify
the following cipctl-options:
ALPHA=˛
specifies the level of significance ˛ for 100.1 ˛/% confidence intervals. The value ˛ must be
between 0 and 1; the default value is 0.05, which results in 95% confidence intervals. The default
value is the value of ALPHA= given in the PROC statement.
LOWERPRE=prefixes
specifies one or more prefixes that are used to create names for variables that contain the lower
confidence limits. To save lower confidence limits for more than one analysis variable, specify a
list of prefixes. The order of the prefixes corresponds to the order of the analysis variables in the
VAR statement.
LOWERNAME=suffixes
specifies one or more suffixes that are used to create names for variables that contain the lower
confidence limits. PROC CAPABILITY creates a variable name by combining the LOWERPRE=
value and suffix name. Because the suffixes are associated with the requested percentiles, list the
suffixes in the same order as the PCTLPTS= percentiles.
TYPE=keyword
specifies the type of confidence limit, where keyword is LOWER, UPPER, SYMMETRIC, or
ASYMMETRIC. The default value is SYMMETRIC.
UPPERPRE=prefixes
specifies one or more prefixes that are used to create names for variables that contain the upper
confidence limits. To save upper confidence limits for more than one analysis variable, specify a
VAR statement.
UPPERNAME=suffixes
specifies one or more suffixes that are used to create names for variables that contain the upper
confidence limits. PROC CAPABILITY creates a variable name by combining the UPPERPRE=
N OTE : See the entries for the PCTLPTS=, PCTLPRE=, and PCTLNAME= options for a detailed
description of how variable names are created using prefixes, percentile values, and suffixes.
CIPCTLNORMAL=(cipctl-options)
CIQUANTNORMAL=(cipctl-options)
requests confidence limits based on the assumption that the data are normally distributed for percentiles
that are requested with the PCTLPTS= option. The computational method is described in Section
4.4.1 of Hahn and Meeker (1991) and uses the noncentral t distribution as given by Odeh and Owen
(1980). This option does not apply if you use a WEIGHT statement. You can specify the following
cipctl-options:
ALPHA=˛
specifies the level of significance ˛ for 100.1 ˛/% confidence intervals. The value ˛ must be
between 0 and 1; the default value is 0.05, which results in 95% confidence intervals. The default
value is the value of ALPHA= given in the PROC statement.
LOWERPRE=prefixes
specifies one or more prefixes that are used to create names for variables that contain the lower
confidence limits. To save lower confidence limits for more than one analysis variable, specify a
VAR statement.
LOWERNAME=suffixes
specifies one or more suffixes that are used to create names for variables that contain the lower
confidence limits. PROC CAPABILITY creates a variable name by combining the LOWERPRE=
TYPE=keyword
default is TWOSIDED.
UPPERPRE=prefixes
specifies one or more prefixes that are used to create names for variables that contain the upper
confidence limits. To save upper confidence limits for more than one analysis variable, specify a
VAR statement.
UPPERNAME=suffixes
specifies one or more suffixes that are used to create names for variables that contain the upper
confidence limits. PROC CAPABILITY creates a variable name by combining the UPPERPRE=
N OTE : See the entries for the PCTLPTS=, PCTLPRE=, and PCTLNAME= options for a detailed
description of how variable names are created using prefixes, percentile values, and suffixes.
PCTLGROUP=BYSTAT | BYVAR
specifies the order in which variables that you request with the PCTLPTS= option are added to the
OUT= data set when the VAR statement lists more than one analysis variable. By default (or if you
specify PCTLGROUP=BYSTAT), all variables that are associated with a percentile value are created
consecutively. If you specify PCTLGROUP=BYVAR, all variables that are associated with an analysis
variable are created consecutively.
Consider the following statements:
proc univariate data=Score;

var PreTest PostTest;
output out=ByStat pctlpts=20 40 pctlpre=Pre_ Post_;
output out=ByVar pctlgroup=byvar pctlpts=20 40 pctlpre=Pre_ Post_;
run;
The order of variables in the data set ByStat is Pre_20, Post_20, Pre_40, Post_40. The order of
variables in the data set ByVar is Pre_20, Pre_40, Post_20, Post_40.
PCTLNAME=suffixes
provides name suffixes for the new variables created by the PCTLPTS= option. These suffixes are
appended to the prefixes you specify with the PCTLPRE= option, replacing the percentile values that
are used as suffixes by default. List the suffixes in the same order in which you specify the percentiles.
If you specify n suffixes with the PCTLNAME= option and m percentile values with the PCTLPTS=
option, where m > n, the suffixes are used to name the first n percentiles, and the default names are
used for the remaining m n percentiles. For example, consider the following statements:
proc capability;
var length width height;
output pctlpts = 20 40
pctlpre = pl pw ph
pctlname = twenty;
run;
The value “twenty” in the PCTLNAME= option is used for only the first percentile in the PCTLPTS=
list. This suffix is appended to the values in the PCTLPRE= option to generate the new variable names
pltwenty, pwtwenty, and phtwenty, which contain the 20th percentiles for length, width, and height,
respectively. Because a second PCTLNAME= suffix is not specified, variable names for the 40th
percentiles for length, width, and height are generated using the prefixes and percentile values. Thus,
the output data set contains the variables pltwenty, pl40, pwtwenty, pw40, phtwenty, and ph40.
PCTLNDEC=value
specifies the number of decimal places in percentile values that are incorporated into percentile variable
names. The default value is 1. For example, the following statements create two output data sets, each
containing one percentile variable. The variable in data set short is named pwid85_1, while the one in
data set long is named pwid85_125.
proc capability;
var width;
output out=short pctlpts=85.125 pctlpre=pwid;
output out=long pctlpts=85.125 pctlpre=pwid pctlndec=3;
run;
PCTLPRE=prefixes
specifies prefixes used to create variable names for percentiles requested with the PCTLPTS= option.
The PCTLPRE= and PCTLPTS= options must be used together.
The procedure generates new variable names by using the prefix and the percentile values. If the
specified percentile is an integer, the variable name is simply the prefix followed by the value. For
noninteger percentiles, an underscore replaces the decimal point in the variable name, and decimal
values are truncated to one decimal place. For example, the following statements create the variables
pwid20, pwid33_3, pwid66_6, and pwid80 for the 20th, 33.33rd, 66.67th, and 80th percentiles of width,
respectively:

var width;
output pctlpts=20 33.33 66.67 80 pctlpre=pwid;
run;
If you request percentiles for more than one variable, you should list prefixes in the same order in
which the variables appear in the VAR statement. For example, the following statements compute the
80th and 87.5th percentiles for length and width and save the new variables plength80, plength87_5,
pwidth80, and pwidth87_5 in the output data set:

var length width;
output pctlpts=80 87.5 pctlpre=plength pwidth;
run;
PCTLPTS=percentiles
specifies percentiles that are not automatically computed by the procedure. The CAPABILITY
procedure automatically computes the 1st, 5th, 10th, 25th, 50th, 75th, 90th, 95th, and 99th percentiles
for the data. These can be saved in an output data set by using keyword=names specifications. The
PCTLPTS= option generates additional percentiles and outputs them to a data set; these additional
percentiles are not printed.
If you use the PCTLPTS= option, you must also use the PCTLPRE= option to provide a prefix for
the new variable names. For example, to create variables that contain the 20th, 40th, 60th, and 80th
percentiles of length, use the following statements:

var length;
output pctlpts=20 40 60 80 pctlpre=plen;
run;
This creates the variables plen20, plen40, plen60, and plen80, whose values are the corresponding
percentiles of length. In addition to specifying name prefixes with the PCTLPRE= option, you can also
use the PCTLNAME= option to create name suffixes for the new variables created by the PCTLPTS=
option.
Details: OUTPUT Statement F 449
Details: OUTPUT Statement

OUT= Data Set
The CAPABILITY procedure creates an OUT= data set for each OUTPUT statement. The new data set
contains an observation for each combination of levels of the variables in the BY and CLASS statements, or
a single observation if you do not specify a BY or CLASS statement. Thus, the number of observations in
the new data set corresponds to the number of groups for which statistics are calculated. The variables in the
new data set are as follows:
variables in the BY statement. The values of these variables match the values in the corresponding BY
group in the DATA= data set.
variables in the CLASS statement. The values of these variables identify the CLASS level within a BY
group in the DATA= data set that from which statistics are computed.
variables in the ID statement. The values of these variables match those for the first observation in
each BY group, or for the first observation in the data set if you do not specify a BY statement.
variables created by selecting statistics in the OUTPUT statement. The values of the statistics are
computed using all the nonmissing data, or statistics are computed for each BY group if you use a BY
statement.
variables created by requesting new percentiles with the PCTLPTS= option. The names of these new
variables depend on the values of the PCTLPRE= and PCTLNAME= options.
If the output data set contains a percentile variable or a quartile variable, the percentile definition assigned
with the PCTLDEF= option in the PROC CAPABILITY statement is recorded on the output data set label.
The values of variables requested with the statistics keywords CP, CPK, CPL, CPM, CPU, K, PCTGTR, and
PCTLSS are missing unless you identify specification limits in a SPEC statement or in a SPEC= data set.
As an alternative to OUT= data sets, you can create an OUTTABLE= data set. The structure of the
OUTTABLE= data set may be more appropriate when you are computing summary statistics and capability
indices for multiple process variables. See “OUTTABLE= Data Set” on page 232.
Examples: OUTPUT Statement

This section provides additional examples of the OUTPUT statement.
Example 6.19: Computing Nonstandard Capability Indices

N OTE : See Computing Nonstandard Capability Indices in the SAS/QC Sample Library.
In recent years, a number of process capability indices that have been proposed in the research literature are
gradually being introduced in applications. As shown in this example, you can compute such indices in the
DATA step after using the OUTPUT statement in the CAPABILITY procedure to save various summary
statistics.
Hardness measurements (in scaled units) for 50 titanium samples are saved as values of the variable Hardness
in the following SAS data set:
data Titanium;
label Hardness = 'Hardness Measurement';
input Hardness @@;
datalines;
1.38 1.49 1.43 1.60 1.59
1.34 1.44 1.64 1.83 1.57
1.45 1.74 1.61 1.39 1.63
1.73 1.61 1.35 1.51 1.47
1.46 1.41 1.56 1.40 1.58
1.43 1.53 1.53 1.58 1.62
1.58 1.46 1.26 1.57 1.41
1.53 1.36 1.63 1.36 1.66
1.49 1.55 1.67 1.41 1.39
1.75 1.37 1.36 1.86 1.49
;
The target value for hardness is 1.6, and the lower and upper specification limits are 0.8 and 2.4, respectively.
The samples are produced by an in-control process, and the measurements are assumed to be normally
distributed.
The following statements use the OUTPUT statement to save various descriptive statistics and an estimate of
the index Cpm in a data set named Indices:
proc capability data=Titanium noprint;

var Hardness;
specs lsl=0.8 target=1.6 usl=2.4;
output out=Indices
n = n
mean = avg
std = std
var = var
lsl = lsl
target = t
usl = usl
pnormal = pnormal
cpm = cpm ;
run;
In addition to Cpm , you want to report an estimate for the index Cpmk , which is defined as follows:
d j mj
Cpmk D p
3 2 C . T /2
where d D .USL LSL/=2, m D .USL C LSL/=2, and and are the mean and standard deviation of the
normal distribution. Refer to Section 3.6 of Kotz and Johnson (1993). A natural estimator for Cpmk is
d jX mj
CO pmk D q P
3 n1 niD1 .Xi T /2
The following statements compute this estimate:

Examples: OUTPUT Statement F 451
data Indices;
set Indices;
d = 0.5*( USL - LSL );
m = 0.5*( USL + LSL );
num = d - abs( avg - m );
den = 3 * sqrt( (n-1)*var/n + (avg-t)*(avg-t) );
cpmk = num/den;
run;
title "Capability Analysis of Titanium Hardness";

proc print data=Indices noobs;
var n avg std lsl t usl cpm cpmk pnormal;
run;
The results are listed in Output 6.19.1.
Output 6.19.1 Computation of Cpmk

Capability Analysis of Titanium Hardness
n avg std lsl t usl cpm cpmk pnormal

50 1.5212 0.13295 0.8 1.6 2.4 1.72545 1.56713 0.25111
Note that the p-value for the Kolmogorov-Smirnov test of normality is 0.25111, indicating that the assumption
of normality is justified.
The following statements also compute an estimate of the index Cpm by using the SPECIALINDICES option:
proc capability data=Titanium specialindices;

var Hardness;
specs lsl=0.8 target=1.6 usl=2.4;
run;
Output 6.19.2 Computation of Cpmk by Using the SPECIALINDICES Option

Capability Analysis of Titanium Hardness

Variable: Hardness (Hardness Measurement)

Cp 2.005745 1.609575 2.401129
CPL 1.808179 1.438675 2.175864
CPU 2.203311 1.757916 2.646912
Cpk 1.808179 1.438454 2.177904
Cpm 1.725446 1.410047 2.066027
Example 6.20: Approximate Confidence Limits for Cpk

N OTE : See Approximate Confidence Limits for Cpk in the SAS/QC Sample Library.
This example illustrates how you can use the OUTPUT statement to compute confidence limits for the
capability index Cpk .
You can request the approximate confidence limits given by Bissell (1990) with the keywords CPKLCL
and CPKUCL in the OUTPUT statement. However, this is not the only method that has been proposed
for computing confidence limits for Cpk . Zhang, Stenback, and Wardrop (1990), referred to here as ZSW,
proposed approximate confidence limits of the form
b pk ˙ kb
C pk
where b pk is an estimator of the standard deviation of C

b pk . Equation (8) of ZSW provides an approximation
to the variance of Cb pk from which one can obtain 100 % confidence limits for Cpk as
2 s 3
1 n 1 .n 1/ 2 ..n 2/=2/ 5
LCL D C
b pk 41 ˆ ..1 /=2/
n 3 2 2 ..n 1/=2/
2 s 3
1 n 1 .n 1/ 2 ..n 2/=2/ 5
UCL D C
b pk 41 C ˆ .1 .1 /=2/
n 3 2 2 ..n 1/=2/
This assumes that Cb pk is normally distributed. You can also compute approximate confidence limits based
on equation (6) of ZSW, which provides an exact expression for the variance of C
b pk .
The following program uses the methods of Bissell (1990) and ZSW to compute approximate confidence
limits for Cpk for the variable Hardness in the data set Titanium (see Example 6.19).
proc capability data=Titanium noprint;

var Hardness;
specs lsl=0.8 usl=2.4;
output out=Summary
n = n
mean = mean
std = std
lsl = lsl
usl = usl
cpk = cpk
cpklcl = cpklcl
cpkucl = cpkucl
cpl = cpl
cpu = cpu ;
data Summary;
set Summary;
length Method $ 16;
Examples: OUTPUT Statement F 453
Method = "Bissell";
lcl = cpklcl;
ucl = cpkucl;
output;
* Assign confidence level;

level = 0.95;
aux = probit( 1 - (1-level)/2 );
Method = "ZSW Equation 6";

zsw = log(0.5*n-0.5)
+ ( 2*(lgamma(0.5*n-1)-lgamma(0.5*n-0.5)) );
zsw = sqrt((n-1)/(n-3)-exp(zsw));
lcl = cpk*(1-aux*zsw);
ucl = cpk*(1+aux*zsw);
output;
Method = "ZSW Equation 8";

ds = 3*(cpu+cpl)/2;
ms = 3*(cpl-cpu)/2;
f1 = (1/3)*sqrt((n-1)/2)*gamma((n-2)/2)*(1/gamma((n-1)/2));
f2 = sqrt(2/n)*(1/gamma(0.5))*exp(-n*0.5*ms*ms);
f3 = ms*(1-(2*probnorm(-sqrt(n)*ms)));
ex = f1*(ds-f2-f3);
sd = ((n-1)/(9*(n-3)))*(ds**2-(2*ds*(f2+f3))+ms**2+(1/n));
sd = sd-(ex*ex);
sd = sqrt(sd);
lcl = cpk-aux*sd;
ucl = cpk+aux*sd;
output;
run;
title "Approximate 95% Confidence Limits for Cpk";

proc print data = Summary noobs;
var Method lcl cpk ucl;
run;
The results are shown in Output 6.20.1.
Output 6.20.1 Approximate Confidence Limits for Cpk

Approximate 95% Confidence Limits for Cpk
Method lcl cpk ucl

Bissell 1.43845 1.80818 2.17790
ZSW Equation 6 1.43596 1.80818 2.18040
ZSW Equation 8 1.42419 1.80818 2.19217
Note that there is fairly close agreement in the three methods.

You can display the confidence limits computed using Bissell’s approach on plots produced by the CAPA-
BILITY procedure by specifying the keywords CPKLCL and CPKUCL in the INSET statement.
The following statements also compute an estimate of the index Cpk along with approximate limits by using
the SPECIALINDICES option:
proc capability data=Titanium specialindices;

var Hardness;
specs lsl=0.8 usl=2.4;
run;
Output 6.20.2 Approximate Confidence Limits for Cpk using the SPECIALINDICES option
Approximate 95% Confidence Limits for Cpk

Variable: Hardness (Hardness Measurement)

Cp 2.005745 1.609575 2.401129
CPL 1.808179 1.438675 2.175864
CPU 2.203311 1.757916 2.646912
Cpk 1.808179 1.438454 2.177904
PPPLOT Statement: CAPABILITY Procedure
Overview: PPPLOT Statement

The PPPLOT statement creates a probability-probability plot (also referred to as a P-P plot or percent plot),
which compares the empirical cumulative distribution function (ecdf) of a variable with a specified theoretical
cumulative distribution function such as the normal. If the two distributions match, the points on the plot form
a linear pattern that passes through the origin and has unit slope. Thus, you can use a P-P plot to determine
how well a theoretical distribution models a set of measurements.
You can specify one of the following theoretical distributions with the PPPLOT statement:
beta
exponential
gamma
Gumbel
inverse Gaussian
lognormal
normal
generalized Pareto
power function
Getting Started: PPPLOT Statement F 455
Rayleigh
Weibull
You can use options in the PPPLOT statement to do the following:
specify or estimate parameters for the theoretical distribution
You can also create a comparative P-P plot by using the PPPLOT statement in conjunction with a CLASS
statement.
You have three alternatives for producing P-P plots with the PPPLOT statement:
statement.
N OTE : Probability-probability plots should not be confused with probability plots, which compare a set of
ordered measurements with percentiles from a specified distribution. You can create probability plots with
the PROBPLOT statement.
Getting Started: PPPLOT Statement

The following example illustrates the basic syntax of the PPPLOT statement. For complete details of the
PPPLOT statement, see the section “Syntax: PPPLOT Statement” on page 457.
Creating a Normal Probability-Probability Plot

N OTE : See Creating P-P Plots in the SAS/QC Sample Library.
The distances between two holes cut into 50 steel sheets are measured and saved as values of the variable
Distance in the following data set: 3
data Sheets;
input Distance @@;
label Distance='Hole Distance in cm';
datalines;
9.80 10.20 10.27 9.70 9.76
10.11 10.24 10.20 10.24 9.63
9.99 9.78 10.10 10.21 10.00
3 These data are also used to create Q-Q plots in “QQPLOT Statement: CAPABILITY Procedure” on page 508.
9.96 9.79 10.08 9.79 10.06

10.10 9.95 9.84 10.11 9.93
10.56 10.47 9.42 10.44 10.16
10.11 10.36 9.94 9.77 9.36
9.89 9.62 10.05 9.72 9.82
9.99 10.16 10.58 10.70 9.54
10.31 10.07 10.33 9.98 10.15
;
The cutting process is in statistical control. As a preliminary step in a capability analysis of the process, it is
decided to check whether the distances are normally distributed. The following statements create a P-P plot,
shown in Figure 6.30, which is based on the normal distribution with mean D 10 and standard deviation
D 0:3:
title 'Normal Probability-Probability Plot for Hole Distance';

proc capability data=Sheets noprint;
ppplot Distance / normal(mu=10 sigma=0.3)
square
odstitle=title;
run;
The NORMAL option in the PPPLOT statement requests a P-P plot based on the normal cumulative
distribution function, and the MU= and SIGMA= normal-options specify and . Note that a P-P plot is
always based on a completely specified distribution, in other words, a distribution with specific parameters.
In this example, if you did not specify the MU= and SIGMA= normal-options, the sample mean and sample
standard deviation would be used for and .
Syntax: PPPLOT Statement F 457
Figure 6.30 Normal P-P Plot with Diagonal Reference Line
The linearity of the pattern in Figure 6.30 is evidence that the measurements are normally distributed with
mean 10 and standard deviation 0.3. The SQUARE option displays the plot in a square format.
Syntax: PPPLOT Statement

The syntax for the PPPLOT statement is as follows:
PPPLOT < variables > < / options > ;
You can specify the keyword PP as an alias for PPPLOT, and you can use any number of PPPLOT statements
in the CAPABILITY procedure. The components of the PPPLOT statement are described as follows.
variables
are the process variables for which to create P-P plots. If you specify a VAR statement, the variables
must also be listed in the VAR statement. Otherwise, the variables can be any numeric variables in the
input data set. If you do not specify a list of variables, then by default, the procedure creates a P-P plot
for each variable listed in the VAR statement or for each numeric variable in the input data set if you
do not specify a VAR statement. For example, each of the following PPPLOT statements produces two
P-P plots, one for length and one for width:
proc capability data=measures;

var length width;
ppplot;
run;

ppplot length width;
run;
options
specify the theoretical distribution for the plot or add features to the plot. If you specify more than
one variable, the options apply equally to each variable. Specify all options after the slash (/) in the
PPPLOT statement. You can specify only one option naming a distribution, but you can specify any
number of other options. The distributions available are the beta, exponential, gamma, Gumbel, inverse
Gaussian, lognormal, normal, generalized Pareto, power function, Rayleigh, and Weibull. By default,
the procedure produces a P-P plot based on the normal distribution.
In the following example, the NORMAL, MU= and SIGMA= options request a P-P plot based on the
normal distribution with mean 10 and standard deviation 0.3. The SQUARE option displays the plot in
a square frame, and the CTEXT= option specifies the text color.

ppplot length width / normal(mu=10 sigma=0.3)
square
ctext=blue;
run;
Summary of Options
The following tables list the PPPLOT statement options by function. For complete descriptions, see the
section “Dictionary of Options” on page 462.
Table 6.60 summarizes the options for requesting a specific theoretical distribution.
Table 6.60 Options for Specifying a Theoretical Distribution
Option Description
BETA(beta-options) Specifies beta P-P plot
EXPONENTIAL(exponential-options) Specifies exponential P-P plot
GAMMA(gamma-options) Specifies gamma P-P plot
GUMBEL(Gumbel-options) Specifies Gumbel P-P plot
IGAUSS(iGauss-options) Specifies inverse Gaussian P-P plot
LOGNORMAL(lognormal-options) Specifies lognormal P-P plot
NORMAL(normal-options) Specifies normal P-P plot
PARETO(Pareto-options) Specifies generalized Pareto P-P plot
POWER(power-options) Specifies power function P-P plot
Option Description
RAYLEIGH(Rayleigh-options) Specifies Rayleigh P-P plot
WEIBULL(Weibull-options) Specifies Weibull P-P plot
Table 6.61 summarizes options that specify distribution parameters and control the display of the diagonal
distribution reference line. Specify these options in parentheses after the distribution option. For example,
the following statements use the NORMAL option to request a normal P-P plot:

ppplot length / normal(mu=10 sigma=0.3 color=red);
run;
The MU= and SIGMA= normal-options specify and for the normal distribution, and the COLOR=
normal-option specifies the color for the line.
Option Description
Distribution Reference Line Options
COLOR= Specifies color of distribution reference line
L= Specifies line type of distribution reference line
NOLINE Suppresses the distribution reference line
SYMBOL= Specifies plotting character for line printer plots
W= Specifies width of distribution reference line
Beta-Options
ALPHA= Specifies shape parameter ˛
BETA= Specifies shape parameter ˇ
SIGMA= Specifies scale parameter
THETA= Specifies lower threshold parameter
Exponential-Options
THETA= Specifies threshold parameter
Gamma-Options
Gumbel-Options
MU= Specifies location parameter
IGauss-Options
LAMBDA= Specifies shape parameter
Option Description
MU= Specifies mean
Lognormal-Options
SIGMA= Specifies shape parameter
ZETA= Specifies scale parameter
Normal-Options
MU= Specifies mean
SIGMA= Specifies standard deviation
Pareto-Options
Power-Options
Rayleigh-Options
Weibull-Options
C= Specifies shape parameter c
General Options
Table 6.62 lists options that control the appearance of the plots.
Table 6.62 General PPPLOT Statement Options
Option Description
CONTENTS= Specifies table of contents entry for P-P plot grouping
HREFLABELS= Specifies line labels for HREF= lines
SQUARE Displays P-P plot in square format
Option Description
VREFLABELS= Specifies line labels for VREF= lines
Graphics Options
ANNOTATE= Provides an annotate data set
HMINOR= Specifies number of minor tick marks on horizontal axis
axis
VMINOR= Specifies number of minor tick marks on vertical axis
VREFLABPOS= Specifies position for VREF= line labels

ODSFOOTNOTE= Specifies footnote displayed on P-P plot
ODSFOOTNOTE2= Specifies secondary footnote displayed on P-P plot
ODSTITLE= Specifies title displayed on P-P plot
ODSTITLE2= Specifies secondary title displayed on P-P plot

cell only
CFRAMESIDE= Specifies color for filling row label frames
CFRAMETOP= Specifies color for filling column label frames
Option Description
INTERTILE= Specifies distance between tiles in comparative plot
NCOLS= Specifies number of columns in comparative plot
NROWS= Specifies number of rows in comparative plot
Options for Line Printer Charts

NOOBSLEGEND Suppresses legend for hidden points
PPSYMBOL= Specifies character for plotted points
The following entries provide detailed descriptions of the options specific to the PPPLOT statement. See
ALPHA=value
specifies the shape parameter ˛ (˛ > 0) for P-P plots requested with the BETA, GAMMA, PARETO,
and POWER options. For examples, see the entries for the distribution options.
BETA< (beta-options) >
creates a beta P-P plot. To create the plot, the n nonmissing observations are ordered from smallest to
largest:
x.1/ x.2/ x.n/
The y-coordinate of the ith point is the empirical cdf value ni . The x-coordinate is the theoretical beta
cdf value
R
x x /˛ 1 .C t /ˇ 1
B˛ˇ .i/ D .i/ .t B.˛;ˇ / .˛Cˇ 1/
dt
.˛/.ˇ /
where B˛ˇ ./ is the normalized incomplete beta function, B.˛; ˇ/ D .˛Cˇ /
, and
D lower threshold parameter

˛ D first shape parameter .˛ > 0/
ˇ D second shape parameter .ˇ > 0/
You can specify ˛, ˇ, , and with the ALPHA=, BETA=, SIGMA=, and THETA= beta-options, as
illustrated in the following example:

ppplot width / beta(theta=1 sigma=2 alpha=3 beta=4);
run;
If you do not specify values for these parameters, then by default, D 0, D 1, and maximum
likelihood estimates are calculated for ˛ and ˇ.
IMPORTANT: If the default unit interval (0,1) does not adequately describe the range of your data,
then you should specify THETA= and SIGMA= so that your data fall in the interval .; C /.
If the data are beta distributed with parameters ˛, ˇ, , and , then the points on the plot for ALPHA=˛,
BETA=ˇ, SIGMA=, and THETA= tend to fall on or near the diagonal line y D x, which is displayed
by default. Agreement between the diagonal line and the point pattern is evidence that the specified
beta distribution is a good fit. You can specify the SCALE= option as an alias for the SIGMA= option
and the THRESHOLD= option as an alias for the THETA= option.
BETA=value
specifies the shape parameter ˇ (ˇ > 0) for P-P plots requested with the BETA distribution option. See
the preceding entry for the BETA distribution option for an example.
C=value
specifies the shape parameter c (c > 0) for P-P plots requested with the WEIBULL option. See the
entry for the WEIBULL option for examples.

EXP< (exponential-options) >
creates an exponential P-P plot. To create the plot, the n nonmissing observations are ordered from
smallest to largest:
x.1/ x.2/ x.n/
i
The y-coordinate of the ith point is the empirical cdf value n. The x-coordinate is the theoretical
exponential cdf value

x.i/
F .x.i / / D 1 exp
where
You can specify and with the SIGMA= and THETA= exponential-options, as illustrated in the
following example:

ppplot width / exponential(theta=1 sigma=2);
run;
If you do not specify values for these parameters, then by default, D 0 and a maximum likelihood
estimate is calculated for .
IMPORTANT: Your data must be greater than or equal to the lower threshold . If the default D 0
is not an adequate lower bound for your data, specify with the THETA= option.
If the data are exponentially distributed with parameters and , the points on the plot for SIGMA=
and THETA= tend to fall on or near the diagonal line y D x, which is displayed by default.
Agreement between the diagonal line and the point pattern is evidence that the specified exponential
distribution is a good fit. You can specify the SCALE= option as an alias for the SIGMA= option and
the THRESHOLD= option as an alias for the THETA= option.

creates a gamma P-P plot. To create the plot, the n nonmissing observations are ordered from smallest
to largest:
x.1/ x.2/ x.n/
i
gamma cdf value
R ˛ 1
x x 1 t t
G˛ .i/ D .i/ .˛/ exp dt
where G˛ ./ is the normalized incomplete gamma function, and
You can specify ˛, , and with the ALPHA=, SIGMA=, and THETA= gamma-options, as illustrated
in the following example:

ppplot width / gamma(alpha=1 sigma=2 theta=3);
run;
If you do not specify values for these parameters, then by default, D 0 and maximum likelihood
estimates are calculated for ˛ and .
is not an adequate lower bound for your data, specify with the THETA= option.
If the data are gamma distributed with parameters ˛, , and , the points on the plot for ALPHA=˛,
SIGMA= , and THETA= tend to fall on or near the diagonal line y D x, which is displayed by
default. Agreement between the diagonal line and the point pattern is evidence that the specified
gamma distribution is a good fit. You can specify the SHAPE= option as an alias for the ALPHA=
option, the SCALE= option as an alias for the SIGMA= option, and the THRESHOLD= option as an
alias for the THETA= option.

creates a Gumbel P-P plot. To create the plot, the n nonmissing observations are ordered from smallest
to largest:
x.1/ x.2/ x.n/
i
Gumbel cdf value

F .x.i / / D exp e .x.i/ /=
where
You can specify and with the MU= and SIGMA= Gumbel-options. By default, maximum
likelihood estimates are computed for and .
If the data are Gumbel distributed with parameters and , the points on the plot for MU= and
SIGMA= tend to fall on or near the diagonal line y D x, which is displayed by default. Agreement
between the diagonal line and the point pattern is evidence that the specified Gumbel distribution is a
good fit.

creates an inverse Gaussian P-P plot. To create the plot, the n nonmissing observations are ordered
from smallest to largest:
x.1/ x.2/ x.n/
The y-coordinate of the ith point is the empirical cdf value ni . The x-coordinate is the theoretical
inverse Gaussian cdf value
(s ) ( s )
x.i / x.i /

2=
F .x.i / / D ˆ 1 Ce ˆ C1
x.i / x.i /

You can specify known values for and with the MU= and LAMBDA= iGauss-options. By default,
the sample mean is calculated for and a maximum likelihood estimate is computed for and .
If the data are inverse Gaussian distributed with parameters and , the points on the plot for MU=
and LAMBDA= tend to fall on or near the diagonal line y D x, which is displayed by default.
Agreement between the diagonal line and the point pattern is evidence that the specified inverse
Gaussian distribution is a good fit.
LAMBDA=value
specifies the shape parameter ( > 0) for P-P plots requested with the IGAUSS option. Enclose the
LAMBDA= option in parentheses after the IGAUSS distribution keyword. If you do not specify a

LNORM< (lognormal-options) >
creates a lognormal P-P plot. To create the plot, the n nonmissing observations are ordered from
x.1/ x.2/ x.n/
i
lognormal cdf value

log.x.i/ /
ˆ
where ˆ./ is the cumulative standard normal distribution function, and
D scale parameter
You can specify , , and with the THETA=, ZETA=, and SIGMA= lognormal-options, as illustrated
in the following example:

ppplot width / lognormal(theta=1 zeta=2);
run;
If you do not specify values for these parameters, then by default, D 0 and estimates of and are
computed as described in “Lognormal Distribution” on page 353.
IMPORTANT: Your data must be greater than the lower threshold . If the default D 0 is not an
adequate lower bound for your data, specify with the THETA= option.
If the data are lognormally distributed with parameters , , and , the points on the plot for SIGMA= ,
THETA= , and ZETA= tend to fall on or near the diagonal line y D x, which is displayed by default.
Agreement between the diagonal line and the point pattern is evidence that the specified lognormal
distribution is a good fit. You can specify the SHAPE= option as an alias for the SIGMA= option, the
SCALE= option as an alias for the ZETA= option, and the THRESHOLD= option as an alias for the
THETA= option.
MU=value
specifies the parameter for a P-P plot requested with the GUMBEL, IGAUSS, and NORMAL options.
For examples, see Figure 6.30, or Figure 6.31 and Figure 6.32. For the normal and inverse Gaussian
distributions, the default value of is the sample mean. If you do not specify a value for for the
Gumbel distribution, the procedure calculates a maximum likelihood estimate.
NOLINE
suppresses the diagonal reference line.
NOOBSLEGEND
NOOBSL
suppresses the legend that indicates the number of hidden observations in a legacy line printer plot.
This option is ignored unless you specify the LINEPRINTER option in the PROC CAPABILITY
statement.

NORM< (normal-options ) >
creates a normal P-P plot. By default, if you do not specify a distribution option, the procedure displays
a normal P-P plot. To create the plot, the n nonmissing observations are ordered from smallest to
largest:
x.1/ x.2/ x.n/
i
normal cdf value

x R x.i/ 1 .t /2
ˆ .i/ D 1 p exp 2 2 dt
2
where ˆ./ is the cumulative standard normal distribution function, and
D location parameter or mean

D scale parameter or standard deviation . > 0/
You can specify and with the MU= and SIGMA= normal-options, as illustrated in the following
example:

ppplot width / normal(mu=1 sigma=2);
run;
By default, the sample mean and sample standard deviation are used for and .
If the data are normally distributed with parameters and , the points on the plot for MU= and
between the diagonal line and the point pattern is evidence that the specified normal distribution is a
good fit. For an example, see Figure 6.30.

creates a generalized Pareto P-P plot. To create the plot, the n nonmissing observations are ordered
from smallest to largest:
x.1/ x.2/ x.n/
i
generalized Pareto cdf value
˛1
˛.x.i / /

F .x.i / / D 1 1

where
The parameter for the generalized Pareto distribution must be less than the minimum data value.
You can specify with the THETA= Pareto-option. The default value for is 0. In addition, the
generalized Pareto distribution has a shape parameter ˛ and a scale parameter . You can specify
these parameters with the ALPHA= and SIGMA= Pareto-options. By default, maximum likelihood
estimates are computed for ˛ and .
If the data are generalized Pareto distributed with parameters , , and ˛, the points on the plot for
THETA=, SIGMA= , and ALPHA=˛ tend to fall on or near the diagonal line y D x, which is
displayed by default. Agreement between the diagonal line and the point pattern is evidence that the
specified generalized Pareto distribution is a good fit.

creates a power function P-P plot. To create the plot, the n nonmissing observations are ordered from
x.1/ x.2/ x.n/
The y-coordinate of the ith point is the empirical cdf value ni . The x-coordinate is the theoretical power
function cdf value
x.i / ˛

F .x.i / / D

where

The power function distribution is bounded below by the parameter and above by the value C .
You can specify and by using the THETA= and SIGMA= power-options. The default values for
and are 0 and 1, respectively.
You can specify a value for the shape parameter, ˛, with the ALPHA= power-option. If you do not
specify a value for ˛, the procedure calculates a maximum likelihood estimate.
The power function distribution is a special case of the beta distribution with its second shape parameter,
ˇ D 1.
If the data are power function distributed with parameters , , and ˛, the points on the plot for
THETA=, SIGMA= , and ALPHA=˛ tend to fall on or near the diagonal line y D x, which is
displayed by default. Agreement between the diagonal line and the point pattern is evidence that the
specified power function distribution is a good fit.
PPSYMBOL=‘character ’
specifies the character used to plot the points in a legacy line printer plot. The default is the plus sign
(+). This option is ignored unless you specify the LINEPRINTER option in the PROC CAPABILITY
statement.

creates a Rayleigh P-P plot. To create the plot, the n nonmissing observations are ordered from smallest
to largest:
x.1/ x.2/ x.n/
i
Rayleigh cdf value
.x.i/ /2 =.2 2 /
F .x.i / / D 1 e
where
The parameter for the Rayleigh distribution must be less than the minimum data value. You can
specify with the THETA= Rayleigh-option. The default value for is 0. You can specify with the
SIGMA= Rayleigh-option. By default, a maximum likelihood estimate is computed for .
If the data are Rayleigh distributed with parameters and , the points on the plot for THETA= and
between the diagonal line and the point pattern is evidence that the specified Rayleigh distribution is a
good fit.
SIGMA=value
specifies the parameter , where > 0. When used with the BETA, EXPONENTIAL, GAMMA,
GUMBEL, NORMAL, PARETO, POWER, RAYLEIGH, and WEIBULL options, the SIGMA= option
specifies the scale parameter. When used with the LOGNORMAL option, the SIGMA= option specifies
the shape parameter. Enclose the SIGMA= option in parentheses after the distribution keyword. For an
example of the SIGMA= option used with the NORMAL option, see Figure 6.30.
SQUARE
displays the P-P plot in a square frame. The default is a rectangular frame. See Figure 6.30 for an
example.
specifies the character used for the diagonal reference line in legacy line printer plots. The default
character is the first letter of the distribution option keyword. This option is ignored unless you specify
the LINEPRINTER option in the PROC CAPABILITY statement.
THETA=value
THRESHOLD=value
specifies the lower threshold parameter for plots requested with the BETA, EXPONENTIAL,
GAMMA, LOGNORMAL, PARETO, POWER, RAYLEIGH, and WEIBULL options.

WEIB< (Weibull-options) >
creates a Weibull P-P plot. To create the plot, the n nonmissing observations are ordered from smallest
to largest:
x.1/ x.2/ x.n/
i
Weibull cdf value
x.i/ c

F .x.i / / D 1 exp
where
c = shape parameter (c > 0)
You can specify c, , and with the C=, SIGMA=, and THETA= Weibull-options, as illustrated in the
following example:

ppplot width / weibull(theta=1 sigma=2);
run;
If you do not specify values for these parameters, then by default D 0 and maximum likelihood
estimates are calculated for and c.
is not an adequate lower bound for your data, you should specify with the THETA= option.
If the data are Weibull distributed with parameters c, , and , the points on the plot for C=c,
SIGMA= , and THETA= tend to fall on or near the diagonal line y D x, which is displayed by
default. Agreement between the diagonal line and the point pattern is evidence that the specified
Weibull distribution is a good fit. You can specify the SHAPE= option as an alias for the C= option,
the SCALE= option as an alias for the SIGMA= option, and the THRESHOLD= option as an alias for
the THETA= option.
ZETA=value
specifies a value for the scale parameter for lognormal P-P plots requested with the LOGNORMAL
option.
Details: PPPLOT Statement

construction and interpretation of P-P plots
comparison of P-P plots with Q-Q plots
distributions supported by the PPPLOT statement
graphical enhancements of P-P plots
Construction and Interpretation of P-P Plots

A P-P plot compares the empirical cumulative distribution function (ecdf) of a variable with a specified
theoretical cumulative distribution function F ./. The ecdf, denoted by Fn .x/, is defined as the proportion of
nonmissing observations less than or equal to x, so that Fn .x.i / / D ni .
To construct a P-P plot, the n nonmissing values are first sorted in increasing order:
x.1/ x.2/ x.n/
Then the ith ordered value x.i / is represented on the plot by the point whose x-coordinate is F .x.i/ / and
whose y-coordinate is ni .
Like Q-Q plots and probability plots, P-P plots can be used to determine how well a theoretical distribution
models a data distribution. If the theoretical cdf reasonably models the ecdf in all respects, including location
and scale, the point pattern on the P-P plot is linear through the origin and has unit slope.
Details: PPPLOT Statement F 471
N OTE : See Interpreting P-P Plots in the SAS/QC Sample Library.

Unlike Q-Q and probability plots, P-P plots are not invariant to changes in location and scale. For example,
the data in the section “Getting Started: PPPLOT Statement” on page 455 are reasonably described by a
normal distribution with mean 10 and standard deviation 0.3. It is instructive to display these data on normal
P-P plots with a different mean and standard deviation, as created by the following statements:
data Sheets;
input Distance @@;
datalines;
9.80 10.20 10.27 9.70 9.76
10.11 10.24 10.20 10.24 9.63
9.99 9.78 10.10 10.21 10.00
9.96 9.79 10.08 9.79 10.06
10.10 9.95 9.84 10.11 9.93
10.56 10.47 9.42 10.44 10.16
10.11 10.36 9.94 9.77 9.36
9.89 9.62 10.05 9.72 9.82
9.99 10.16 10.58 10.70 9.54
10.31 10.07 10.33 9.98 10.15
;

ppplot Distance / normal(mu=9.5 sigma=0.3) square;
ppplot Distance / normal(mu=10 sigma=0.5) square;
run;
Graphics, so the P-P plots are created using ODS Graphics instead of traditional graphics. The resulting plots
are shown in Figure 6.31 and Figure 6.32.
Figure 6.31 Normal P-P Plot with Mean Specified Incorrectly

Figure 6.32 Normal P-P Plot with Standard Deviation Specified Incorrectly
Specifying a mean of 9.5 instead of 10 results in the plot shown in Figure 6.31, while specifying a standard
deviation of 0.5 instead of 0.3 results in the plot shown in Figure 6.32. Both plots clearly reveal the model
misspecification.
Comparison of P-P Plots and Q-Q Plots

A P-P plot compares the empirical cumulative distribution function of a data set with a specified theoretical
cumulative distribution function F ./. A Q-Q plot compares the quantiles of a data distribution with the
quantiles of a standardized theoretical distribution from a specified family of distributions. There are three
important differences in the way P-P plots and Q-Q plots are constructed and interpreted:
The construction of a Q-Q plot does not require that the location or scale parameters of F ./ be specified.
The theoretical quantiles are computed from a standard distribution within the specified family. A
linear point pattern indicates that the specified family reasonably describes the data distribution, and
the location and scale parameters can be estimated visually as the intercept and slope of the linear
pattern. In contrast, the construction of a P-P plot requires the location and scale parameters of F ./ to
evaluate the cdf at the ordered data values.
The linearity of the point pattern on a Q-Q plot is unaffected by changes in location or scale. On a P-P
plot, changes in location or scale do not necessarily preserve linearity.
On a Q-Q plot, the reference line representing a particular theoretical distribution depends on the
location and scale parameters of that distribution, having intercept and slope equal to the location and
scale parameters. On a P-P plot, the reference line for any distribution is always the diagonal line
y D x.
Consequently, you should use a Q-Q plot if your objective is to compare the data distribution with a family of
distributions that vary only in location and scale, particularly if you want to estimate the location and scale
parameters from the plot.
An advantage of P-P plots is that they are discriminating in regions of high probability density, because in
these regions the empirical and theoretical cumulative distributions change more rapidly than in regions of
low probability density. For example, if you compare a data distribution with a particular normal distribution,
differences in the middle of the two distributions are more apparent on a P-P plot than on a Q-Q plot.
For further details on P-P plots, refer to Gnanadesikan (1997) and Wilk and Gnanadesikan (1968).
Summary of Theoretical Distributions

You can use the PPPLOT statement to request P-P plots based on the theoretical distributions summarized in
Table 6.63.
Table 6.63 Distributions and Parameters
Parameters
Family Distribution Function F .x/ Range Location Scale Shape
Rx .t /˛ 1 . C t /ˇ 1
Beta B.˛;ˇ / .˛Cˇ 1/
dt <x < C ˛, ˇ

x
Exponential 1 exp x
Rx ˛ 1
1 t t
Gamma .˛/ exp dt x> ˛

Gumbel exp e .x /= All x
q
x
Inverse Gaussian ˆ x 1 C x>0
q
2= x
e ˆ x C1

Rx
p 1 .log.t / /2
Lognormal 2.t / exp 2 2
dt x>

Rx
p1 .t /2
Normal 1 2 exp 2 2
dt All x
1=˛
˛.x /
Generalized Pareto 1 1 All x ˛
˛
x
Power function <x < C ˛
Parameters
Family Distribution Function F .x/ Range Location Scale Shape
.x /2 =.2 2 /
Rayleigh 1 e x
c
x
Weibull 1 exp x> c
You can request these distributions with the BETA, EXPONENTIAL, GAMMA, GUMBEL, IGAUSS,
NORMAL, LOGNORMAL, PARETO, POWER, RAYLEIGH, and WEIBULL options, respectively. If you
do not specify a distribution option, a normal P-P plot is created.
To create a P-P plot, you must provide all of the parameters for the theoretical distribution. If you do not
specify parameters, then default values or estimates are substituted, as summarized by Table 6.64.
Table 6.64 Defaults for Parameters
Family Default Values Estimated Values

Beta D 0, D 1 Maximum likelihood estimates for ˛
and ˇ
Exponential D0 Maximum likelihood estimate for
Gamma D0 Maximum likelihood estimates for
and ˛
Gumbel None Maximum likelihood estimates for
and
Inverse Gaussian None Sample estimate for , maximum
likelihood estimate for
Lognormal D0 Estimates for and that are
computed as described in
“Lognormal Distribution” on
page 353
Normal None Sample estimates for and
Generalized Pareto D0 Maximum likelihood estimates for
and ˛
Power function D 0, D 1 Maximum likelihood estimate for ˛
Rayleigh D0 Maximum likelihood estimate for
Weibull D0 Maximum likelihood estimates for
and c
Specification of Symbol Markers

If you produce traditional graphics, you can use options in the SYMBOL1 statement to specify the appearance
of the symbol marker for the points. The V= option specifies the symbol, the C= option specifies the color,
and the H= option specifies the height. Refer to SAS/GRAPH: Reference for details concerning these options.
If you produce a line printer plot, you can use the PPSYMBOL= option in the PPPLOT statement to specify
the character used to plot the points.
Specification of the Distribution Reference Line

If you produce traditional graphics, you can control the color, type, and width of the diagonal distribution
reference line by specifying the COLOR=, L=, and W= options in parentheses after the distribution option in
the PPPLOT statement. Alternatively, you can control these features with the C=, L=, and W= options in the
SYMBOL4 statement. Refer to SAS/GRAPH: Reference for details concerning these options. If you produce
a line printer plot, you can specify the character used for the line with the SYMBOL= option enclosed in
parentheses after the distribution option in the PPPLOT statement.
ODS Graphics
destination where the graph is produced. PPPLOT options used to control the appearance of traditional
When ODS Graphics is in effect, the PPPLOT statement assigns a name to the graph it creates. You can use
this name to reference the graph when using ODS. The name is listed in Table 6.65.
Table 6.65 ODS Graphics Produced by the PPPLOT Statement

PPPlot P-P plot
producing charts.
PROBPLOT Statement: CAPABILITY Procedure F 477
PROBPLOT Statement: CAPABILITY Procedure
Overview: PROBPLOT Statement

The PROBPLOT statement creates a probability plot, which compares ordered values of a variable with
percentiles of a specified theoretical distribution such as the normal. If the data distribution matches the
theoretical distribution, the points on the plot form a linear pattern. Thus, you can use a probability plot to
determine how well a theoretical distribution models a set of measurements.
You can specify one of the following theoretical distributions with the PROBPLOT statement:
beta
exponential
gamma
Gumbel
three-parameter lognormal
normal
generalized Pareto
power function
Rayleigh
two-parameter Weibull
three-parameter Weibull
You can use options in the PROBPLOT statement to do the following:
specify or estimate shape parameters for the theoretical distribution
display a reference line corresponding to specified or estimated location and scale parameters for the
theoretical distribution
You can also create a comparative probability plot by using the PROBPLOT statement in conjunction with a
CLASS statement.
You have three alternatives for producing probability plots the PROBPLOT statement:
statement.
N OTE : Probability plots are similar to Q-Q plots, which you can create with the QQPLOT statement (see
“QQPLOT Statement: CAPABILITY Procedure” on page 508). Probability plots are preferable for graphical
estimation of percentiles, whereas Q-Q plots are preferable for graphical estimation of distribution parameters
and capability indices.
Getting Started: PROBPLOT Statement

The following examples illustrate the basic syntax of the PROBPLOT statement. For complete details of the
PROBPLOT statement, see the section “Syntax: PROBPLOT Statement” on page 483. Advanced examples
are provided on the section “Examples: PROBPLOT Statement” on page 505.
Creating a Normal Probability Plot

N OTE : See Creating a Normal Probability Plot in the SAS/QC Sample Library.
The diameters of 50 steel rods are measured and saved as values of the variable Diameter in the following
data set:4
data Rods;
input Diameter @@;
label Diameter='Diameter in mm';
datalines;
5.501 5.251 5.404 5.366 5.445
5.576 5.607 5.200 5.977 5.177
5.332 5.399 5.661 5.512 5.252
5.404 5.739 5.525 5.160 5.410
5.823 5.376 5.202 5.470 5.410
5.394 5.146 5.244 5.309 5.480
5.388 5.399 5.360 5.368 5.394
5.248 5.409 5.304 6.239 5.781
5.247 5.907 5.208 5.143 5.304
5.603 5.164 5.209 5.475 5.223
;
The process producing the rods is in statistical control, and as a preliminary step in a capability analysis of
the process, you decide to check whether the diameters are normally distributed. The following statements
create the normal probability plot shown in Figure 6.33:
title 'Normal Probability Plot for Diameters';

proc capability data=Rods noprint;
probplot Diameter / odstitle=title;
4 This data set is analyzed using quantile-quantile plots in Example 6.23 and Example 6.24.
Getting Started: PROBPLOT Statement F 479
run;
Figure 6.33 Normal Probability Plot Created with Traditional Graphics
Note that the PROBPLOT statement creates a normal probability plot for Diameter by default.
The nonlinearity of the point pattern indicates a departure from normality. Because the point pattern is
curved with slope increasing from left to right, a theoretical distribution that is skewed to the right, such as a
lognormal distribution, should provide a better fit than the normal distribution. This possibility is explored in
the next example.
Creating Lognormal Probability Plots

N OTE : See Creating Lognormal Probability Plots in the SAS/QC Sample Library.
When you request a lognormal probability plot, you must specify the shape parameter for the lognormal
distribution (see Table 6.70 for the equation). The value of must be positive, and typical values of range
from 0.1 to 1.0. Alternatively, you can specify that is to be estimated from the data.
The following statements illustrate the first approach by creating a series of three lognormal probability plots
for the variable Diameter introduced in the preceding example:

probplot Diameter / lognormal(sigma=0.2 0.5 0.8)
href = 95
square;
run;
The LOGNORMAL option requests plots based on the lognormal family of distributions, and the SIGMA=
option requests plots for equal to 0.2, 0.5, and 0.8. The SQUARE option displays the probability plot in a
square format and the HREF= option requests a reference line at the 95th percentile.
The resulting plots are displayed in Figure 6.34, Figure 6.35, and Figure 6.36, respectively. The value
D 0:5 in Figure 6.35 produces the most linear pattern.
Figure 6.34 Probability Plot Based on Lognormal Distribution with D 0:2

Getting Started: PROBPLOT Statement F 481

Based on Figure 6.35, the 95th percentile of the diameter distribution is approximately 5.9 mm, because this
is the value corresponding to the intersection of the point pattern with the reference line.
The following statements illustrate how you can create a lognormal probability plot for Diameter by using an
estimate of . (See “Lognormal Distribution” on page 353 for a description of how the estimate is computed.)
title 'Lognormal Probability Plot for Diameters';

probplot Diameter / lognormal(sigma=est)
href = 95
odstitle = title
square;
run;
The plot is displayed in Figure 6.37.
Note that the estimate of (in this case 0.041) does not necessarily produce the most linear point pattern.
This example is continued in Example 6.22.
Figure 6.37 Probability Plot Based on Lognormal Distribution with Estimated

Syntax: PROBPLOT Statement F 483
Syntax: PROBPLOT Statement

The syntax for the PROBPLOT statement is as follows:
PROBPLOT < variables > < / options > ;
You can specify the keyword PROB as an alias for PROBPLOT, and you can use any number of PROBPLOT
statements in the CAPABILITY procedure. The components of the PROBPLOT statement are described as
follows.
variables
are the process variables for which to create probability plots. If you specify a VAR statement, the
variables must also be listed in the VAR statement. Otherwise, the variables can be any numeric
variables in the input data set. If you do not specify a list of variables, then by default the procedure
creates a probability plot for each variable listed in the VAR statement, or for each numeric variable
in the DATA= data set if you do not specify a VAR statement. For example, each of the following
PROBPLOT statements produces two probability plots, one for length and one for width:

var length width;
probplot;
run;

probplot length width;
run;
options
one variable, the options apply equally to each variable. Specify all options after the slash (/) in the
PROBPLOT statement. You can specify only one option naming the distribution in each PROBPLOT
statement, but you can specify any number of other options. The distributions available are the beta,
exponential, gamma, Gumbel, lognormal, normal, generalized Pareto, power function, Rayleigh,
two-parameter Weibull, and three-parameter Weibull. By default, the procedure produces a plot for the
normal distribution.
In the following example, the NORMAL option requests a normal probability plot for each variable,
while the MU= and SIGMA= normal-options request a distribution reference line corresponding to
the normal distribution with D 10 and D 0:3. The SQUARE option displays the plot in a square
frame, and the CTEXT= option specifies the text color.

probplot length1 length2 / normal(mu=10 sigma=0.3)
square
ctext=blue;
run;
Summary of Options
The following tables list the PROBPLOT statement options by function. For complete descriptions, see the
section “Dictionary of Options” on page 488.
Option Description
BETA(beta-options) Specifies beta probability plot for shape
parameters ˛, ˇ specified with mandatory
ALPHA= and BETA= beta-options
EXPONENTIAL(exponential-options) Specifies exponential probability plot
GAMMA(gamma-options) Specifies gamma probability plot for
shape parameter ˛ specified with
mandatory ALPHA= gamma-option
GUMBEL(Gumbel-options) Specifies Gumbel probability plot
LOGNORMAL(lognormal-options) Specifies lognormal probability plot for
shape parameter specified with
mandatory SIGMA= lognormal-option
NORMAL(normal-options) Specifies normal probability plot
PARETO(Pareto-options) Specifies generalized Pareto probability
plot for shape parameter ˛ specified with
mandatory ALPHA= Pareto-option
POWER(power-options) Specifies power function probability plot
for shape parameter ˛ specified with
mandatory ALPHA= power-option
RAYLEIGH(Rayleigh-options) Specifies Rayleigh probability plot
WEIBULL(Weibull-options) Specifies three-parameter Weibull
probability plot for shape parameter c
specified with mandatory C=
Weibull-option
WEIBULL2(Weibull2-options) Specifies two-parameter Weibull
probability plot
Table 6.67 summarizes options that specify distribution parameters and control the display of a distribution
reference line. Specify these options in parentheses after the distribution option. For example, the following
statements use the NORMAL option to request a normal probability plot with a distribution reference line:

probplot length / normal(mu=10 sigma=0.3 color=red);
run;
The MU= and SIGMA= normal-options display a distribution reference line that corresponds to the normal
distribution with mean 0 D 10 and standard deviation 0 D 0:3, and the COLOR= normal-option specifies
the color for the line.
Option Description
Beta-Options
ALPHA= Specifies mandatory shape parameter ˛
BETA= Specifies mandatory shape parameter ˇ
SIGMA= Specifies 0 for distribution reference line
THETA= Specifies 0 for distribution reference line
Exponential-Options
Gamma-Options
Gumbel-Options
MU= Specifies location parameter
Lognormal-Options
SIGMA= Specifies mandatory shape parameter
SLOPE= Specifies slope of distribution reference line
ZETA= Specifies 0 for distribution reference line (slope is exp.0 /)
Normal-Options
MU= Specifies 0 for distribution reference line
Pareto-Options
Power-Options
Option Description
Rayleigh-Options
Weibull-Options
C= Specifies mandatory shape parameter c
Weibull2-Options
C= Specifies c0 for distribution reference line (slope is 1=c0 )
SIGMA= Specifies 0 for distribution reference line (intercept is
log.0 /)
SLOPE= Specifies slope of distribution reference line
THETA= Specifies known lower threshold 0
General Options
Table 6.68 General PROBPLOT Statement Options
Option Description
CONTENTS= Specifies table of contents entry for probability plot grouping
GRID Draws grid lines perpendicular to the percentile axis
HREFLABELS= Specifies line labels for HREF= lines
LEGEND= Identifies LEGEND statement
NADJ= Adjusts sample size (N) when computing percentiles
NOLINELEGEND Suppresses distribution reference line information in legend
NOSPECLEGEND Suppresses specifications information in legend
PCTLMINOR Requests minor tick marks for percentile axis
PCTLORDER= Specifies tick mark labels for percentile axis
RANKADJ= Adjusts ranks when computing percentiles
ROTATE Switches horizontal and vertical axes
SQUARE Displays plot in square format
VREFLABELS= Specifies line labels for VREF= lines
Option Description
Graphics Options
axis

ODSFOOTNOTE= Specifies footnote displayed on probability plot
ODSFOOTNOTE2= Specifies secondary footnote displayed on probability plot
ODSTITLE= Specifies title displayed on probability plot
ODSTITLE2= Specifies secondary title displayed on probability plot
Option Description
ANNOKEY Applies annotation to key cell only
NCOLS= Specifies number of columns in comparative probability plot
NROWS= Specifies number of rows in comparative probability plot

GRIDCHAR= Specifies character for GRID lines
HREFCHAR= Specifies character for HREF= lines
PROBSYMBOL= Specifies character for plotted points
VREFCHAR= Specifies character for VREF= lines
The following sections provide detailed descriptions of options specific to the PROBPLOT statement. See
General Options
graphics:
ALPHA=value-list|EST
specifies values for a mandatory shape parameter ˛ .˛ > 0/ for probability plots requested with
the BETA, GAMMA, PARETO, and POWER options. A plot is created for each value specified.
For examples, see the entries for the distribution options. If you specify ALPHA=EST, a maximum
likelihood estimate is computed for ˛.
BETA(ALPHA=value-list|EST BETA=value-list|EST < beta-options >)

creates a beta probability plot for each combination of the shape parameters ˛ and ˇ given by the
mandatory ALPHA= and BETA= options. If you specify ALPHA=EST and BETA=EST, a plot is
created based on maximum likelihood estimates for ˛ and ˇ. In the following examples, the first
PROBPLOT statement produces one plot, the second statement produces four plots, the third statement
produces six plots, and the fourth statement produces one plot:

probplot width / beta(alpha=2 beta=2);
probplot width / beta(alpha=2 3 beta=1 2);
probplot width / beta(alpha=2 to 3 beta=1 to 2 by 0.5);
probplot width / beta(alpha=est beta=est);
run;
To create the plot, the observations are ordered

from smallest to largest, and the ith ordered observation
1 i 0:375
is plotted against the quantile B˛ˇ nC0:25 , where B˛ˇ1 ./ is the inverse normalized incomplete beta
function, n is the number of nonmissing observations, and ˛ and ˇ are the shape parameters of the
beta distribution. The horizontal axis is scaled in percentile units.
The point pattern on the plot for ALPHA=˛ and BETA=ˇ tends to be linear with intercept and slope
if the data are beta distributed with the specific density function
(
.x /˛ 1 . C x/ˇ 1
B.˛;ˇ / .˛Cˇ 1/
for < x < C
p.x/ D
0 for x or x C
.˛/.ˇ /
and

The intercept and slope are based on the quantile scale for the horizontal axis, which is displayed on a
Q-Q plot; see “QQPLOT Statement: CAPABILITY Procedure” on page 508.
To obtain graphical estimates of ˛ and ˇ, specify lists of values for the ALPHA= and BETA= options,
and select the combination of ˛ and ˇ that most nearly linearizes the point pattern.
To assess the point pattern, you can add a diagonal distribution reference line corresponding to 0 and
0 with the beta-options THETA=0 and SIGMA=0 . Alternatively, you can add a line corresponding
to estimated values of 0 and 0 with the beta-options THETA=EST and SIGMA=EST. Specify these
options in parentheses, as in the following example:

probplot width / beta(alpha=2 beta=3 theta=4 sigma=5);
run;
Agreement between the reference line and the point pattern indicates that the beta distribution with
parameters ˛, ˇ, 0 and 0 is a good fit. You can specify the SCALE= option as an alias for the
SIGMA= option and the THRESHOLD= option as an alias for the THETA= option.
BETA=value-list|EST
specifies values for the shape parameter ˇ .ˇ > 0/ for probability plots requested with the BETA
distribution option. A plot is created for each value specified with the BETA= option. If you specify
BETA=EST, a maximum likelihood estimate is computed for ˇ. For examples, see the preceding entry
for the BETA option.
C=value(-list)|EST
specifies the shape parameter c (c > 0) for probability plots requested with the WEIBULL and
WEIBULL2 options. You must specify C= as a Weibull-option with the WEIBULL option; in this
situation it accepts a list of values, or if you specify C=EST, a maximum likelihood estimate is computed
for c. You can optionally specify C=value or C=EST as a Weibull2-option with the WEIBULL2 option
to request a distribution reference line; in this situation, you must also specify SIGMA=value or
SIGMA=EST.
For example, the first PROBPLOT statement below creates three three-parameter Weibull plots
corresponding to the shape parameters c = 1, c = 2, and c = 3. The second PROBPLOT statement
creates a single three-parameter Weibull plot corresponding to an estimated value of c. The third
PROBPLOT statement creates a single two-parameter Weibull plot with a distribution reference line
corresponding to c0 D 2 and 0 D 3.

probplot width / weibull(c=1 2 3);
probplot width / weibull(c=est);
probplot width / weibull2(c=2 sigma=3);
run;

EXP(< exponential-options >)
creates an exponential probability plot. To create the plot, the observations are ordered from smallest
i 0:375
to largest, and the ith ordered observation is plotted against the quantile log 1 nC0:25 , where n
is the number of nonmissing observations. The horizontal axis is scaled in percentile units.
The point pattern on the plot tends to be linear with intercept and slope if the data are exponentially
distributed with the specific density function
(
1 x
exp for x
p.x/ D
0 for x <
where is a threshold parameter, and is a positive scale parameter.
To assess the point pattern, you can add a diagonal distribution reference line corresponding to 0
and 0 with the exponential-options THETA=0 and SIGMA=0 . Alternatively, you can add a
line corresponding to estimated values of 0 and 0 with the exponential-options THETA=EST and
SIGMA=EST. Specify these options in parentheses, as in the following example:

probplot width / exponential(theta=4 sigma=5);
run;
Agreement between the reference line and the point pattern indicates that the exponential distribution
with parameters 0 and 0 is a good fit. You can specify the SCALE= option as an alias for the
GAMMA(ALPHA=value-list|EST < gamma-options > )

creates a gamma probability plot for each value of the shape parameter ˛ given by the mandatory
ALPHA= option. If you specify ALPHA=EST, a plot is created based on a maximum likelihood
estimate for ˛.
For example, the first PROBPLOT statement below creates three plots corresponding to ˛ D 0:4,
˛ D 0:5, and ˛ D 0:6. The second PROBPLOT statement creates a single plot.

probplot width / gamma(alpha=0.4 to 0.6 by 0.2);
probplot width / gamma(alpha=est);
run;
To create the plot, the observations are

orderedfrom smallest to largest, and the ith ordered observation
is plotted against the quantile G˛ inC0:25
1 0:375
, where G˛ 1 ./ is the inverse normalized incomplete
gamma function, n is the number of nonmissing observations, and ˛ is the shape parameter of the
gamma distribution. The horizontal axis is scaled in percentile units.
The point pattern on the plot for ALPHA=˛ tends to be linear with intercept and slope if the data
are gamma distributed with the specific density function
( ˛ 1
1 x x
exp for x >
p.x/ D .˛/
0 for x
where
To obtain a graphical estimate of ˛, specify a list of values for the ALPHA= option, and select the
value that most nearly linearizes the point pattern.
To assess the point pattern, you can add a diagonal distribution reference line corresponding to 0 and 0
with the gamma-options THETA=0 and SIGMA=0 . Alternatively, you can add a line corresponding
to estimated values of 0 and 0 with the gamma-options THETA=EST and SIGMA=EST. Specify
these options in parentheses, as in the following example:

probplot width / gamma(alpha=2 theta=3 sigma=4);
run;
Agreement between the reference line and the point pattern indicates that the gamma distribution with
parameters ˛, 0 and 0 is a good fit. You can specify the SCALE= option as an alias for the SIGMA=
option and the THRESHOLD= option as an alias for the THETA= option.
GRID
draws reference lines perpendicular to the percentile axis at major tick marks.
GUMBEL(< Gumbel-options >)

creates a Gumbel probability plot. To create the plot, the observations areordered
from smallest
to
i 0:375
largest, and the ith ordered observation is plotted against the quantile log log nC0:25 , where
n is the number of nonmissing observations. The horizontal axis is scaled in percentile units.
The point pattern on the plot tends to be linear with intercept and slope if the data are Gumbel
e .x /=
.x /=
p.x/ D exp e

where is a location parameter and is a positive scale parameter.
0 with the Gumbel-options MU=0 and SIGMA=0 . Alternatively, you can add a line corresponding
to estimated values of 0 and 0 with the Gumbel-options MU=EST and SIGMA=EST. Specify these
options in parentheses following the GUMBEL option.
Agreement between the reference line and the point pattern indicates that the Gumbel distribution with
parameters 0 and 0 is a good fit.
LOGNORMAL(SIGMA=value-list|EST < lognormal-options >)

LNORM(SIGMA=value-list|EST < lognormal-options >)
creates a lognormal probability plot for each value of the shape parameter given by the mandatory
SIGMA= option or its alias, the SHAPE= option. If you specify SIGMA=EST, a plot is created based
on an estimate of that is computed as described in “Lognormal Distribution” on page 353.
For example, the first PROBPLOT statement below produces two plots, and the second PROBPLOT
statement produces a single plot:

probplot width / lognormal(sigma=1.5 2.5 l=2);
probplot width / lognormal(sigma=est);
run;

fromsmallest to largest, and the ith ordered observation
1 i 0:375
is plotted against the quantile exp ˆ nC0:25 , where ˆ 1 ./ is the inverse standard cumulative
normal distribution, n is the number of nonmissing observations, and is the shape parameter of the
lognormal distribution. The horizontal axis is scaled in percentile units.
The point pattern on the plot for SIGMA= tends to be linear with intercept and slope exp./ if the
data are lognormally distributed with the specific density function

.log.x / /2
(
p 1 exp for x >
p.x/ D 2.x / 2 2
0 for x
where
D scale parameter
To obtain a graphical estimate of , specify a list of values for the SIGMA= option, and select the
with the lognormal-options THETA=0 and ZETA=0 . Alternatively, you can add a line corresponding
to estimated values of 0 and 0 with the lognormal-options THETA=EST and ZETA=EST.
Specify these options in parentheses, as in the following example:

probplot width / lognormal(sigma=2 theta=3 zeta=0);
run;
Agreement between the reference line and the point pattern indicates that the lognormal distribution
with parameters , 0 , and 0 is a good fit. See Example 6.22 for an example.
You can specify the THRESHOLD= option as an alias for the THETA= option and the SCALE= option
as an alias for the ZETA= option.
MU=value|EST
specifies the mean 0 for a probability plot requested with the GUMBEL and NORMAL options. If
you specify MU=EST, 0 is equal to the sample mean for the normal distribution. For the Gumbel
distribution, a maximum likelihood estimate is calculated. See Example 6.21.
NADJ=value
specifies the adjustment value added to the sample size in the calculation of theoretical percentiles.
The default is 14 , as recommended by Blom (1958). Also refer to Chambers et al. (1983) for additional
information.
NOLEGEND
suppresses legends for specification limits, fitted curves, distribution lines, and hidden observations.
NOLINELEGEND
NOLINEL
suppresses the legend for the optional distribution reference line.

NORM< (normal-options) >
creates a normal probability plot. This is the default if you do not specify a distribution option. To
create the plot, the observations are
ordered from smallest to largest, and the ith ordered observation is
1 i 0:375 1
plotted against the quantile ˆ nC0:25 , where ˆ ./ is the inverse cumulative standard normal
distribution, and n is the number of nonmissing observations. The horizontal axis is scaled in percentile
units.
The point pattern on the plot tends to be linear with intercept and slope if the data are normally
distributed with the specific

.x /2
p.x/ D p1 exp 2 2 for all x
2
where is the mean and is the standard deviation ( > 0).

0 with the normal-options MU=0 and SIGMA=0 . Alternatively, you can add a line corresponding
to estimated values of 0 and 0 with the normal-options MU=EST and SIGMA=EST; the estimates
of 0 and 0 are the sample mean and sample standard deviation.
Specify these options in parentheses, as in the following example:

probplot length / normal(mu=10 sigma=0.3);
probplot length / normal(mu=est sigma=est);
run;
Agreement between the reference line and the point pattern indicates that the normal distribution with
NOSPECLEGEND
NOSPECL
suppresses the legend for specification limit reference lines.
PARETO(< Pareto-options >)

creates a generalized Pareto probability plot for each value of the shape parameter ˛ given by the
mandatory ALPHA= option. If you specify ALPHA=EST, a plot is created based on a maximum
likelihood estimate for ˛.
To create the plot, the observations are ordered from smallest to largest, and the ith ordered observation
is plotted against the quantile .1 .1 inC0:250:375 ˛
/ /=˛ (˛ ¤ 0) or log.1 inC0:25 0:375
/ (˛ D 0), where
n is the number of nonmissing observations and ˛ is the shape parameter of the generalized Pareto
distribution. The horizontal axis is scaled in percentile units.
are generalized Pareto distributed with the specific density function
.1 ˛.x /= /1=˛ 1 if ˛ ¤ 0

1
p.x/ D 1
exp. .x /= / if ˛ D 0
where

with the Pareto-options THETA=0 and SIGMA=0 . Alternatively, you can add a line corresponding
to estimated values of 0 and 0 with the Pareto-options THETA=EST and SIGMA=EST. Specify
these options in parentheses following the PARETO option.
Agreement between the reference line and the point pattern indicates that the generalized Pareto
distribution with parameters ˛, 0 , and 0 is a good fit.
PCTLORDER=value-list
specifies the tick mark values labeled on the theoretical percentile axis. Because the values are
percentiles, the labels must be between 0 and 100, exclusive. The values must be listed in increasing
order and must cover the plotted percentile range. Otherwise, a default list is used. For example,
consider the following:

probplot length / pctlorder=1 10 25 50 75 90 99;
run;
Note that the ORDER= option in the AXIS statement is not supported by the PROBPLOT statement.
POWER(< power-options >)

creates a power function probability plot for each value of the shape parameter ˛ given by the mandatory
estimate for ˛.
To create the plot, the observations are orderedfrom smallest to largest, and the ith ordered observation
1 i 0:375 1
is plotted against the quantile B˛.1/ nC0:25 , where B˛.1/ ./ is the inverse normalized incomplete
beta function, n is the number of nonmissing observations, ˛ is one shape parameter of the beta
distribution, and the second shape parameter, ˇ D 1. The horizontal axis is scaled in percentile units.
are power function distributed with the specific density function
( ˛ 1
˛ x
for < x < C
p.x/ D
0 for x or x C
where
with the power-options THETA=0 and SIGMA=0 . Alternatively, you can add a line corresponding
to estimated values of 0 and 0 with the power-options THETA=EST and SIGMA=EST. Specify
these options in parentheses following the POWER option.
Agreement between the reference line and the point pattern indicates that the power function distribution
with parameters ˛, 0 , and 0 is a good fit.
RANKADJ=value
specifies the adjustment value added to the ranks in the calculation of theoretical percentiles. The
default is 38 , as recommended by Blom (1958). Also refer to Chambers et al. (1983) for additional
information.
RAYLEIGH(< Rayleigh-options >)

creates a Rayleigh probability plot. To create the plot, the observationsr are ordered from smallest to

largest, and the ith ordered observation is plotted against the quantile 2 log 1 inC0:25
0:375
, where
n is the number of nonmissing observations. The horizontal axis is scaled in percentile units.
The point pattern on the plot tends to be linear with intercept and slope if the data are Rayleigh
x
2
exp. .x /2 =.2 2 // for x
p.x/ D
0 for x <

with the Rayleigh-options THETA=0 and SIGMA=0 . Alternatively, you can add a line corresponding
to estimated values of 0 and 0 with the Rayleigh-options THETA=EST and SIGMA=EST. Specify
these options in parentheses after the RAYLEIGH option.
Agreement between the reference line and the point pattern indicates that the Rayleigh distribution
with parameters 0 and 0 is a good fit.
ROTATE
switches the horizontal and vertical axes so that the theoretical percentiles are plotted vertically while
the data are plotted horizontally. Regardless of whether the plot has been rotated, horizontal axis
options (such as HAXIS=) still refer to the horizontal axis, and vertical axis options (such as VAXIS=)
still refer to the vertical axis. All other options that depend on axis placement adjust to the rotated axes.
SIGMA=value-list|EST
specifies the value of the parameter , where > 0. Alternatively, you can specify SIGMA=EST to
request an estimate of 0 that is computed as
s
Pn
i D1 .log .xi / /
O0 D
n 1
The interpretation and use of the SIGMA= option depend on the distribution option with which it is
specified, as indicated by Table 6.69.
Distribution Option Use of the SIGMA= Option

BETA THETA=0 and SIGMA=0 request a distribution reference
EXPONENTIAL line corresponding to 0 and 0 .
GAMMA
PARETO
POWER
RAYLEIGH
WEIBULL
GUMBEL MU=0 and SIGMA=0 request a distribution reference line
corresponding to 0 and 0 .
LOGNORMAL SIGMA=1 : : : n requests n probability plots with shape
parameters 1 : : : n . The SIGMA= option must be specified.
NORMAL MU=0 and SIGMA=0 request a distribution reference line
corresponding to 0 and 0 . SIGMA=EST requests a line with
0 equal to the sample standard deviation.
WEIBULL2 SIGMA=0 and C=c0 request a distribution reference line
corresponding to 0 and c0 .
In the following example, the first PROBPLOT statement requests a normal plot with a distribution
reference line corresponding to 0 D 5 and 0 D 2, and the second PROBPLOT statement requests a
lognormal plot with shape parameter D 3:

probplot length / normal(mu=5 sigma=2);
probplot width / lognormal(sigma=3);
run;
SLOPE=value|EST
specifies the slope for a distribution reference line requested with the LOGNORMAL and WEIBULL2
options. The intercept and slope are based on the quantile scale for the horizontal axis, which is
displayed on a Q-Q plot; see “QQPLOT Statement: CAPABILITY Procedure” on page 508.
When you use the SLOPE= option with the LOGNORMAL option, you must also specify a threshold
parameter value 0 with the THETA= lognormal-option to request the line. The SLOPE= option is an
alternative to the ZETA= lognormal-option for specifying 0 , because the slope is equal to exp.0 /.
When you use the SLOPE= option with the WEIBULL2 option, you must also specify a scale parameter
value 0 with the SIGMA= Weibull2-option to request the line. The SLOPE= option is an alternative
to the C= Weibull2-option for specifying c0 , because the slope is equal to 1=c0 . See “Location and
Scale Parameters” on page 503.
For example, the first and second PROBPLOT statements below produce the same set of probability
plots as the third and fourth PROBPLOT statements:

probplot width / lognormal(sigma=2 theta=0 zeta=0);
probplot width / weibull2(sigma=2 theta=0 c=0.25);
probplot width / lognormal(sigma=2 theta=0 slope=1);
probplot width / weibull2(sigma=2 theta=0 slope=4);
run;
SQUARE
displays the probability plot in a square frame. For an example, see Output 6.22.1. The default is a
rectangular frame.
THETA=value|EST
THRESHOLD=value
specifies the lower threshold parameter for probability plots requested with the BETA, EXPONEN-
TIAL, GAMMA, LOGNORMAL, PARETO, POWER, RAYLEIGH, WEIBULL, and WEIBULL2
options. When used with the WEIBULL2 option, the THETA= option specifies the known lower
threshold 0 , for which the default is 0. When used with the other distribution options, the THETA=
option specifies 0 for a distribution reference line; alternatively in this situation, you can specify
THETA=EST to request a maximum likelihood estimate for 0 . To request the line, you must also
specify a scale parameter. See Output 6.22.1 for an example of the THETA= option with a lognormal
probability plot.
WEIBULL(C=value-list|EST < Weibull-options >)
WEIB(C=value-list < Weibull-options >)
creates a three-parameter Weibull probability plot for each value of the shape parameter c given by the
mandatory C= option or its alias, the SHAPE= option. If you specify C=EST, a plot is created based
on a maximum likelihood estimate for c. In the following example, the first PROBPLOT statement
creates four plots, and the second PROBPLOT statement creates a single plot:

probplot width / weibull(c=1.8 to 2.4 by 0.2 w=2);
probplot width / weibull(c=est);
run;
1c
is plotted against the quantile log 1 inC0:25
0:375
, where n is the number of nonmissing obser-
vations, and c is the Weibull distribution shape parameter. The horizontal axis is scaled in percentile
units.
The point pattern on the plot for C=c tends to be linear with intercept and slope if the data are
Weibull distributed with the specific density function
( c 1 c
c x x
exp for x >
p.x/ D
0 for x
where
To obtain a graphical estimate of c, specify a list of values for the C= option, and select the value that
most nearly linearizes the point pattern.
with the Weibull-options THETA=0 and SIGMA=0 . Alternatively, you can add a line corresponding
to estimated values of 0 and 0 with the Weibull-options THETA=EST and SIGMA=EST. Specify

probplot width / weibull(c=2 theta=3 sigma=4);
run;
Agreement between the reference line and the point pattern indicates that the Weibull distribution with
parameters c, 0 , and 0 is a good fit. You can specify the SCALE= option as an alias for the SIGMA=
WEIBULL2< (Weibull2-options) >

W2< (Weibull2-options) >
creates a two-parameter Weibull probability plot. You should use the WEIBULL2 option when your
data have a known lower threshold 0 . You can specify the threshold value 0 with the THETA=
Weibull2-option or its alias, the THRESHOLD= Weibull2-option. The default is 0 D 0.
To create the plot, the observations are ordered from smallest to largest, and the log of the
shifted
ith ordered observation
x.i / , denoted by log.x.i / 0 /, is plotted against the quantile
i 0:375
log log 1 nC0:25 , where n is the number of nonmissing observations. The horizontal axis
is scaled in percentile units. Note that the C= shape parameter option is not mandatory with the
WEIBULL2 option.
1
The point pattern on the plot for THETA=0 tends to be linear with intercept log. / and slope c if the
data are Weibull distributed with the specific density function
( c 1 c
c x 0 x 0
exp for x > 0
p.x/ D
0 for x 0
where
0 D known lower threshold


An advantage of the two-parameter Weibull plot over the three-parameter Weibull plot is that the
parameters c and can be estimated from the slope and intercept of the point pattern. A disadvantage
is that the two-parameter Weibull distribution applies only in situations where the threshold parameter
is known.
To assess the point pattern, you can add a diagonal distribution reference line corresponding to 0
and c0 with the Weibull2-options SIGMA=0 and C=c0 . Alternatively, you can add a distribution
reference line corresponding to estimated values of 0 and c0 with the Weibull2-options SIGMA=EST
and C=EST. Specify these options in parentheses, as in the following example:

probplot width / weibull2(theta=3 sigma=4 c=2);
run;
Agreement between the distribution reference line and the point pattern indicates that the Weibull
distribution with parameters c0 , 0 and 0 is a good fit. You can specify the SCALE= option as an
alias for the SIGMA= option and the SHAPE= option as an alias for the C= option.
ZETA=value|EST
specifies a value for the scale parameter for lognormal probability plots requested with the LOG-
NORMAL option. Specify THETA=0 and ZETA=0 to request a distribution reference line with
intercept 0 and slope exp.0 /. See Output 6.22.1 for an example.

CGRID=color
specifies the color for the grid lines requested by the GRID option.
LEGEND=name | NONE
and fitted curves. Specifying LEGEND=NONE is equivalent to specifying the NOLEGEND option.
LGRID=linetype
specifies the line type for the grid lines requested by the GRID option.
PCTLMINOR
requests minor tick marks for the percentile axis. See Output 6.22.1 for an example.
WGRID=n
specifies the width of the grid lines requested with the GRID option. If you use the WGRID= option,
you do not need to specify the GRID option.

Details: PROBPLOT Statement F 501
GRIDCHAR=‘character ’
specifies the character used for the lines requested by the GRID option for a line printer plot. The
default is the vertical bar (|).
NOOBSLEGEND
NOOBSL
suppresses the legend that indicates the number of hidden observations.
PROBSYMBOL=‘character ’
specifies the character used to mark the points in a line printer plot. The default is the plus sign (+).
specifies the character used to display the distribution reference line in a line printer plot. The default
character is the first letter of the distribution option keyword.
Details: PROBPLOT Statement

distributions supported by the PROBPLOT statement
SYMBOL statement options

You can use the PROBPLOT statement to request probability plots based on the theoretical distributions
summarized in Table 6.70.
Table 6.70 Distributions and Parameters
Parameters
Distribution Density Function p.x/ Range Location Scale Shape
.x /˛ 1 .C x/ˇ 1
<x < C ˛, ˇ

1 x
Exponential exp x
˛ 1
1 x x
Gamma .˛/ exp x> ˛
.x /=

e
Gumbel exp e .x /= all x

p 1 .log.x / /2
Lognormal exp 2 2 x>
2.x /
(3-parameter)

p1 .x /2
Normal exp 2 2
all x
2
Parameters
Generalized ˛¤0 1
.1 ˛.x /= /1=˛ 1
x> ˛
1
Pareto ˛D0 exp. .x /= /
˛ 1
˛ x
Power function x> ˛
x
Rayleigh 2
exp. .x /2 =.2 2 // x
c 1 c
c x x
Weibull exp x> c
(3-parameter)
c 1 c
c x 0 x 0
Weibull exp x > 0 0 c
(2-parameter) (known)
You can request these distributions with the BETA, EXPONENTIAL, GAMMA, LOGNORMAL, NORMAL,
WEIBULL, and WEIBULL2 options, respectively. If you do not specify a distribution option, a normal
probability plot is created.
Shape Parameters
Some of the distribution options in the PROBPLOT statement require you to specify one or two shape
parameters in parentheses after the distribution keyword. These are summarized in Table 6.71.
Table 6.71 Shape Parameter Options for the

PROBPLOT Statement
Distribution Keyword Mandatory Shape Parameter Option Range

BETA ALPHA=˛, BETA=ˇ ˛ > 0, ˇ > 0
EXPONENTIAL None
GAMMA ALPHA=˛ ˛>0
GUMBEL None
LOGNORMAL SIGMA= >0
NORMAL None
PARETO ALPHA=˛ ˛>0
POWER ALPHA=˛ ˛>0
RAYLEIGH None
WEIBULL C=c c>0
WEIBULL2 None
You can visually estimate the value of a shape parameter by specifying a list of values for the shape parameter
option. The PROBPLOT statement produces a separate plot for each value. You can then use the value of the
shape parameter producing the most nearly linear point pattern. Alternatively, you can request that the plot
Details: PROBPLOT Statement F 503
be created using an estimated shape parameter. For an example, see “Creating Lognormal Probability Plots”
on page 479.
Location and Scale Parameters

If you specify the location and scale parameters for a distribution (or if you request estimates for these
parameters), a diagonal distribution reference line is displayed on the plot. (An exception is the two-
parameter Weibull distribution, for which a line is displayed when you specify or estimate the scale and
shape parameters.) Agreement between this line and the point pattern indicates that the distribution with
these parameters is a good fit. For illustrations, see Example 6.21 and Example 6.22.
Table 6.72 shows how the specified parameters determine the intercept5 and slope of the line:
Table 6.72 Intercept and Slope of Distribution Reference Line
Parameters Linear Pattern

Distribution Location Scale Shape Intercept Slope
Beta ˛,ˇ
Exponential
Gamma ˛
Gumbel
Lognormal exp./
Normal
Generalized Pareto ˛
Power function ˛
Rayleigh
Weibull (3-parameter) c
1
Weibull (2-parameter) 0 (known) c log. / c
For the LOGNORMAL and WEIBULL2 options, you can specify the slope directly with the SLOPE= option.
That is, for the LOGNORMAL option, specifying THETA=0 and SLOPE=exp.0 / displays the same line as
specifying THETA=0 and ZETA=0 . For the WEIBULL2 option, specifying SIGMA=0 and SLOPE= c10
displays the same line as specifying SIGMA=0 and C=c0 .
SYMBOL Statement Options

In earlier releases of SAS/QC software, graphical features of lower and upper specification lines and diagonal
distribution reference lines were controlled with options in the SYMBOL2, SYMBOL3, and SYMBOL4
statements, respectively. These options are still supported, although they have been superseded by options
in the PROBPLOT and SPEC statements. Table 6.73 summarizes the two sets of options. N OTE : These
statements have no effect on ODS Graphics output.
5 The intercept and slope are based on the quantile scale for the horizontal axis, which is displayed on a Q-Q plot; see “QQPLOT
Statement: CAPABILITY Procedure” on page 508.

Table 6.73 SYMBOL Statement Options

Symbol markers SYMBOL1 statement
Character VALUE=special-symbol
Color COLOR=color
Font FONT=font
Height HEIGHT=value
Position LSL=value
Color CLSL=color COLOR=color
Line type LLSL=linetype LINE=linetype
Width WLSL=value WIDTH=value
Position USL=value
Color CUSL=color COLOR=color
Line type LUSL=linetype LINE=linetype
Width WUSL=value WIDTH=value
Target reference line SPEC statement
Color CTARGET=color
Line type LTARGET=linetype
Width WTARGET=value
Distribution reference line PROBPLOT statement SYMBOL4 statement
Color COLOR=color COLOR=color
Line type LINE=linetype LINE=linetype
Width WIDTH=value WIDTH=value
For an illustration of these options, see Example 6.21.
ODS Graphics
destination where the graph is produced. PROBPLOT options used to control the appearance of traditional
When ODS Graphics is in effect, the PROBPLOT statement assigns a name to the graph it creates. You can
use this name to reference the graph when using ODS. The name is listed in Table 6.74.
Examples: PROBPLOT Statement F 505
Table 6.74 ODS Graphics Produced by the PROBPLOT Statement

ProbPlot Probability plot
producing charts.
Examples: PROBPLOT Statement

This section provides advanced examples of the PROBPLOT statement.
Example 6.21: Displaying a Normal Reference Line

N OTE : See Probability Plot with Normal Reference Line in the SAS/QC Sample Library.
Measurements of the distance between two holes cut into 50 steel sheets are saved as values of the variable
Distance in the following data set:
data Sheets;
input Distance @@;
datalines;
9.80 10.20 10.27 9.70 9.76
10.11 10.24 10.20 10.24 9.63
9.99 9.78 10.10 10.21 10.00
9.96 9.79 10.08 9.79 10.06
10.10 9.95 9.84 10.11 9.93
10.56 10.47 9.42 10.44 10.16
10.11 10.36 9.94 9.77 9.36
9.89 9.62 10.05 9.72 9.82
9.99 10.16 10.58 10.70 9.54
10.31 10.07 10.33 9.98 10.15
;
The cutting process is in control, and you decide to check whether the process distribution is normal. The
following statements create a normal probability plot for Distance with lower and upper specification lines at
9.5 cm and 10.5 cm:
title 'Normal Probability Plot for Hole Distance';

probplot Distance / normal(mu=est sigma=est)
square
odstitle=title
nospeclegend;
run;
The plot is shown in Output 6.21.1. The MU= and SIGMA= normal-options request the diagonal reference
line that corresponds to the normal distribution with estimated parameters O D 10:027 and O D 0:2889. The
LSL= and USL= SPEC statement options request the lower and upper specification lines. The SYMBOL
statement specifies the symbol marker for the plotted points.
Output 6.21.1 Normal Reference Line
Example 6.22: Displaying a Lognormal Reference Line

N OTE : See Creating Lognormal Probability Plots in the SAS/QC Sample Library.
This example is a continuation of “Creating Lognormal Probability Plots” on page 479. Figure 6.35 shows
that a lognormal distribution with shape parameter D 0:5 is a good fit for the distribution of Diameter in
the data set Rods.
The lognormal distribution involves two other parameters: a threshold parameter and a scale parameter .
See Table 6.70 for the equation of the lognormal density function. The following statements illustrate how
you can request a diagonal distribution reference line whose slope and intercept are determined by estimates
of and .
Examples: PROBPLOT Statement F 507
title 'Lognormal Probability Plot for Diameters';

probplot Diameter / lognormal(sigma=0.5 theta=est zeta=est)
square
pctlminor
href = 95
hreflabel = '95%'
vref = 5.8 to 6.0 by 0.1
odstitle = title;
run;
The plot is shown in Output 6.22.1.
Output 6.22.1 Lognormal Reference Line
The close agreement between the diagonal reference line and the point pattern indicates that the specific
lognormal distribution with O D 0:5, O D 5:004, and O D 1:003 is a good fit for the diameter measurements.
Specifying HREF=95 adds a reference line indicating the 95th percentile of the lognormal distribution. The
HREFLABEL= option specifies a label for this line. The PCTLMINOR option displays minor tick marks on
the percentile axis. The VREF= option adds reference lines indicating diameter values of 5.8, 5.9, and 6.0,
and the CHREF= and CVREF= options specify colors for the horizontal and vertical reference lines.
Based on the intersection of the diagonal reference line with the HREF= line, the estimated 95th percentile of
the diameter distribution is 5.85 mm.
Note that you could also construct a similar plot in which all three parameters are estimated by substituting
SIGMA=EST for SIGMA=0.5 in the preceding statements.
QQPLOT Statement: CAPABILITY Procedure
Overview: QQPLOT Statement

The QQPLOT statement creates a quantile-quantile plot (Q-Q plot), which compares ordered values of a
variable with quantiles of a specified theoretical distribution such as the normal. If the data distribution
matches the theoretical distribution, the points on the plot form a linear pattern. Thus, you can use a Q-Q plot
to determine how well a theoretical distribution models a set of measurements.
You can specify one of the following theoretical distributions with the QQPLOT statement:
beta
exponential
gamma
Gumbel
three-parameter lognormal
normal
generalized Pareto
power function
Rayleigh
two-parameter Weibull
three-parameter Weibull
You can use options in the QQPLOT statement to do the following:
specify or estimate parameters for the theoretical distribution
display a reference line corresponding to specific location and scale parameters for the theoretical
distribution
Getting Started: QQPLOT Statement F 509
You can also create a comparative Q-Q plot by using the QQPLOT statement in conjunction with a CLASS
statement.
You have three alternatives for producing Q-Q plots with the QQPLOT statement:
statement.
N OTE : Q-Q plots are similar to probability plots, which you can create with the PROBPLOT statement (see
“PROBPLOT Statement: CAPABILITY Procedure” on page 477). Q-Q plots are preferable for graphical
estimation of distribution parameters and capability indices, whereas probability plots are preferable for
graphical estimation of percentiles.
Getting Started: QQPLOT Statement

The following examples illustrate the basic syntax of the QQPLOT statement. For complete details of the
QQPLOT statement, see the section “Syntax: QQPLOT Statement” on page 512. Advanced examples are
provided on the section “Examples: QQPLOT Statement” on page 539.
Creating a Normal Quantile-Quantile Plot

N OTE : See Creating Normal Q-Q Plots in the SAS/QC Sample Library.
Measurements of the distance between two holes cut into 50 steel sheets are saved as values of the variable
Distance in the following data set:
data Sheets;
input Distance @@;
datalines;
9.80 10.20 10.27 9.70 9.76
10.11 10.24 10.20 10.24 9.63
9.99 9.78 10.10 10.21 10.00
9.96 9.79 10.08 9.79 10.06
10.10 9.95 9.84 10.11 9.93
10.56 10.47 9.42 10.44 10.16
10.11 10.36 9.94 9.77 9.36
9.89 9.62 10.05 9.72 9.82
9.99 10.16 10.58 10.70 9.54
10.31 10.07 10.33 9.98 10.15
;
The cutting process is in control, and you decide to check whether the process distribution is normal. The
following statements create a Q-Q plot for Distance, shown in Figure 6.38, with lower and upper specification
lines at 9.5 cm and 10.5 cm:6
title 'Normal Quantile-Quantile Plot for Hole Distance';

qqplot Distance / odstitle=title;
run;
The plot compares the ordered values of Distance with quantiles of the normal distribution. The linearity
of the point pattern indicates that the measurements are normally distributed. Note that a normal Q-Q plot
is created by default. The specification lines are requested with the LSL= and USL= options in the SPEC
statement.
Figure 6.38 Normal Quantile-Quantile Plot Created with Traditional Graphics
6 For a P-P plot using these data, see Figure 6.30. For a probability plot using these data, see Example 6.22.
Getting Started: QQPLOT Statement F 511
Adding a Distribution Reference Line

In a normal Q-Q plot, the normal distribution with mean 0 and standard deviation 0 is represented by a
reference line with intercept 0 and slope 0 . The following statements reproduce the Q-Q plot in Figure 6.38,
adding the line for which 0 and 0 are estimated by the sample mean and standard deviation:

qqplot Distance / normal(mu=est sigma=est)
square
nospeclegend
odstitle=title;
run;
The plot is displayed in Figure 6.39.
Figure 6.39 Adding a Distribution Reference Line to a Q-Q Plot
Specifying MU=EST and SIGMA=EST with the NORMAL option requests the reference line (alternatively,
you can specify numeric values for 0 and 0 with the MU= and SIGMA= options). The COLOR= and L=
options specify the color of the line and the line type. The SQUARE option displays the plot in a square
format, and the NOSPECLEGEND option suppresses the legend for the specification lines.
Syntax: QQPLOT Statement

The syntax for the QQPLOT statement is as follows:
QQPLOT < variables > < / options > ;
You can specify the keyword QQ as an alias for QQPLOT, and you can use any number of QQPLOT
statements in the CAPABILITY procedure. The components of the QQPLOT statement are described as
follows.
variables
are the process variables for which to create Q-Q plots. If you specify a VAR statement, the variables
must also be listed in the VAR statement. Otherwise, the variables can be any numeric variables in the
input data set. If you do not specify a list of variables, then by default the procedure creates a Q-Q
plot for each variable listed in the VAR statement, or for each numeric variable in the DATA= data
set if you do not specify a VAR statement. For example, each of the following QQPLOT statements
produces two Q-Q plots, one for length and one for width:

var length width;
qqplot;
run;

qqplot length width;
run;
options
one variable, the options apply equally to each variable. Specify all options after the slash (/) in
the QQPLOT statement. You can specify only one option naming the distribution in each QQPLOT
statement, but you can specify any number of other options. The distributions available are the beta,
exponential, gamma, Gumbel, lognormal, normal, generalized Pareto, power function, Rayleigh,
two-parameter Weibull, and three-parameter Weibull. By default, the procedure produces a plot for the
normal distribution.
In the following example, the NORMAL option requests a normal Q-Q plot for each variable. The
MU= and SIGMA= normal-options request a distribution reference line with intercept 10 and slope
0.3 for each plot, corresponding to a normal distribution with mean D 10 and standard deviation
D 0:3. The SQUARE option displays the plot in a square frame, and the CTEXT= option specifies
the text color.

qqplot length1 length2 / normal(mu=10 sigma=0.3)
square
Syntax: QQPLOT Statement F 513
ctext=blue;
run;
Summary of Options
The following tables list the QQPLOT statement options by function. For complete descriptions, see
Option Description
BETA(beta-options) Specifies beta Q-Q plot for shape parameters ˛, ˇ
specified with mandatory ALPHA= and BETA=
beta-options
EXPONENTIAL(exponential-options) Specifies exponential Q-Q plot
GAMMA(gamma-options) Specifies gamma Q-Q plot for shape parameter ˛
specified with mandatory ALPHA=
gamma-option
GUMBEL(Gumbel-options) Specifies Gumbel Q-Q plot
LOGNORMAL(lognormal-options) Specifies lognormal Q-Q plot for shape parameter
specified with mandatory SIGMA=
lognormal-option
NORMAL(normal-options) Specifies normal Q-Q plot
PARETO(Pareto-options) Specifies generalized Pareto Q-Q plot for shape
parameter ˛ specified with mandatory ALPHA=
Pareto-option
POWER(power-options) Specifies power function Q-Q plot for shape
parameter ˛ specified with mandatory ALPHA=
power-option
RAYLEIGH(Rayleigh-options) Specifies Rayleigh Q-Q plot
WEIBULL(Weibull-options) Specifies three-parameter Weibull Q-Q plot for
shape parameter c specified with mandatory C=
Weibull-option
WEIBULL2(Weibull2-options) Specifies two-parameter Weibull Q-Q plot
Table 6.76 summarizes options that specify parameter values for theoretical distributions and that control the
display of a distribution reference line. Specify these options in parentheses after the distribution option. For
example, the following statements use the NORMAL option to request a normal Q-Q plot with a specific
distribution reference line. The MU= and SIGMA= normal-options display a distribution reference line with
intercept 10 and slope 0.3. The COLOR= normal-option draws the line in red.

qqplot length / normal(mu=10 sigma=0.3 color=red);
run;
Option Description
Beta-Options
BETA= Specifies mandatory shape parameter ˇ
SIGMA= Specifies reference line slope
THETA= Specifies reference line intercept
Exponential-Options
Gamma-Options
Gumbel-Options
MU= Specifies reference line intercept
Lognormal-Options
SIGMA= Specifies mandatory shape parameter
SLOPE= Specifies reference line slope
ZETA= Specifies reference line slope exp.0 /
Normal-Options
CPKREF Specifies vertical reference lines at intersection of specification
limits with distribution reference line
CPKSCALE Rescales horizontal axis in Cpk units
MU= Specifies reference line intercept
Pareto-Options
Option Description
Power-Options
Rayleigh-Options
Weibull-Options
C= Specifies mandatory shape parameter c
Weibull2-Options
C= Specifies c0 for reference line (slope is c10 )
SIGMA= Specifies 0 for reference line (intercept is log.0 /)
SLOPE= Specifies reference line slope
THETA= Specifies known lower threshold 0
General Options
Table 6.77 General QQPLOT Statement Options
Option Description
CONTENTS= Specifies table of contents entry for Q-Q plot grouping
LEGEND= Specifies LEGEND statement
NADJ= Adjusts sample size (N) when computing quantiles
NOLINELEGEND Suppresses distribution reference line information in legend
NOSPECLEGEND Suppresses specifications information in legend
PCTLAXIS Adds a nonlinear percentile axis
PCTLMINOR Adds minor tick marks to percentile axis
PCTLSCALE Replaces theoretical quantiles with percentiles
Option Description
RANKADJ= Adjusts ranks when computing quantiles
ROTATE Switches horizontal and vertical axes
SQUARE Displays Q-Q plot in square format
Graphics Options
GRID Draws grid lines perpendicular to the quantile axis
WGRID= Specifies thickness for grid lines

ODSFOOTNOTE= Specifies footnote displayed on Q-Q plot
Option Description
ODSFOOTNOTE2= Specifies secondary footnote displayed on Q-Q plot
ODSTITLE= Specifies title displayed on Q-Q plot
ODSTITLE2= Specifies secondary title displayed on Q-Q plot

cell only
NCOLS= Specifies number of columns in comparative Q-Q plot
NROWS= Specifies number of rows in comparative Q-Q plot

QQSYMBOL= Specifies character for plotted points
VREFCHAR= Specifies character for VREF= lines
The following sections provide detailed descriptions of options specific to the QQPLOT statement. See
General Options
graphics:
ALPHA=value-list|EST
specifies values for a mandatory shape parameter ˛ .˛ > 0/ for Q-Q plots requested with the BETA,
GAMMA, PARETO, and POWER options. A plot is created for each value specified. For examples, see
the entries for the distribution options. If you specify ALPHA=EST, a maximum likelihood estimate is
computed for ˛.
BETA(ALPHA=value-list|EST BETA=value-list|EST < beta-options >)

creates a beta Q-Q plot for each combination of the shape parameters ˛ and ˇ given by the mandatory
ALPHA= and BETA= options. If you specify ALPHA=EST and BETA=EST, a plot is created based
on maximum likelihood estimates for ˛ and ˇ. In the following example, the first QQPLOT statement
produces one plot, the second statement produces four plots, the third statement produces six plots,
and the fourth statement produces one plot:

qqplot width / beta(alpha=2 beta=2);
qqplot width / beta(alpha=2 3 beta=1 2);
qqplot width / beta(alpha=2 to 3 beta=1 to 2 by 0.5);
qqplot width / beta(alpha=est beta=est);
run;

from smallest to largest, and the ith ordered observation
1 i 0:375
is plotted against the quantile B˛ˇ nC0:25 , where B˛ˇ1 ./ is the inverse normalized incomplete beta
function, n is the number of nonmissing observations, and ˛ and ˇ are the shape parameters of the
beta distribution.
The point pattern on the plot for ALPHA=˛ and BETA=ˇ tends to be linear with intercept and slope
if the data are beta distributed with the specific density function
(
.x /˛ 1 . C x/ˇ 1
B.˛;ˇ / .˛Cˇ 1/
for < x < C
p.x/ D
0 for x or x C
where B.˛; ˇ/ D .˛/.ˇ
.˛Cˇ /
/
, and
To obtain graphical estimates of ˛ and ˇ, specify lists of values for the ALPHA= and BETA= options,
and select the combination of ˛ and ˇ that most nearly linearizes the point pattern. To assess the
point pattern, you can add a diagonal distribution reference line with intercept 0 and slope 0 with
the beta-options THETA=0 and SIGMA=0 . Alternatively, you can add a line corresponding to
estimated values of 0 and slope 0 with the beta-options THETA=EST and SIGMA=EST. Specify

qqplot width / beta(alpha=2 beta=3 theta=4 sigma=5);
run;
Agreement between the reference line and the point pattern indicates that the beta distribution with
parameters ˛, ˇ, 0 , and 0 is a good fit. You can specify the SCALE= option as an alias for the
BETA=value-list|EST
specifies values for the shape parameter ˇ .ˇ > 0/ for Q-Q plots requested with the BETA distribution
option. A plot is created for each value specified with the BETA= option. If you specify BETA=EST, a
maximum likelihood estimate is computed for ˇ. For examples, see the preceding entry for the BETA
distribution option.
C=value(-list)|EST
specifies the shape parameter c (c > 0) for Q-Q plots requested with the WEIBULL and WEIBULL2
options. You must specify C= as a Weibull-option with the WEIBULL option; in this situation it accepts
a list of values, or if you specify C=EST, a maximum likelihood estimate is computed for c. You can
optionally specify C=value or C=EST as a Weibull2-option with the WEIBULL2 option to request a
distribution reference line; in this situation, you must also specify SIGMA=value or SIGMA=EST. For
an example, see Output 6.25.1.
CPKSCALE
rescales the quantile axis in Cpk units for plots requested with the NORMAL option. Specify
CPKSCALE in parentheses after the NORMAL option. You can use the CPKSCALE option with the
CPKREF option for graphical estimation of the capability indices CPU , CPL, and Cpk , as illustrated
in Output 6.26.1.
EXPONENTIAL(< (exponential-options) >

EXP< (exponential-options) >)
creates an exponential Q-Q plot. To create the plot, the observations are ordered from smallest to
largest, and the ith ordered observation is plotted against the quantile log 1 inC0:25
0:375
, where n is
the number of nonmissing observations.
The pattern on the plot tends to be linear with intercept and slope if the data are exponentially
(
1 x
exp for x
p.x/ D
0 for x <
where is the threshold parameter, and is the scale parameter . > 0/.
To assess the point pattern, you can add a diagonal distribution reference line with intercept 0 and
slope 0 with the exponential-options THETA=0 and SIGMA=0 . Alternatively, you can add a line
corresponding to estimated values of 0 and slope 0 with the exponential-options THETA=EST and
SIGMA=EST. Specify these options in parentheses, as in the following example: as in the following
example:

qqplot width / exponential(theta=4 sigma=5);
run;
Agreement between the reference line and the point pattern indicates that the exponential distribution
with parameters 0 and 0 is a good fit. You can specify the SCALE= option as an alias for the
GAMMA(ALPHA=value-list|EST < gamma-options > )

creates a gamma Q-Q plot for each value of the shape parameter ˛ given by the mandatory ALPHA=
option or its alias, the SHAPE= option. The following example produces three probability plots:

qqplot width / gamma(alpha=0.4 to 0.6 by 0.1);
run;
To create the plot, the observations are

orderedfrom smallest to largest, and the ith ordered observation
is plotted against the quantile G˛ inC0:25
1 0:375
, where G˛ 1 ./ is the inverse normalized incomplete
gamma function, n is the number of nonmissing observations, and ˛ is the shape parameter of the
gamma distribution.
The pattern on the plot for ALPHA=˛ tends to be linear with intercept and slope if the data are
gamma distributed with the specific density function
( ˛ 1
1 x x
exp for x >
p.x/ D .˛/
0 for x
where
To assess the point pattern, you can add a diagonal distribution reference line with intercept 0
and slope 0 with the gamma-options THETA=0 and SIGMA=0 . Alternatively, you can add a
line corresponding to estimated values of 0 and 0 with the gamma-options THETA=EST and
SIGMA=EST. Specify these options in parentheses, as in the following example:

qqplot width / gamma(alpha=2 theta=3 sigma=4);
run;
Agreement between the reference line and the point pattern indicates that the gamma distribution with
parameters ˛, 0 , and 0 is a good fit. You can specify the SCALE= option as an alias for the SIGMA=
GUMBEL(< Gumbel-options >)

creates a Gumbel Q-Q plot. To create the plot, the observations are ordered
from smallest
to largest,
i 0:375
and the ith ordered observation is plotted against the quantile log log nC0:25 , where n is the
number of nonmissing observations.
The point pattern on the plot tends to be linear with intercept and slope if the data are Gumbel
e .x /=
.x /=
p.x/ D exp e

where is a location parameter and is a positive scale parameter.

0 with the Gumbel-options MU=0 and SIGMA=0 . Alternatively, you can add a line corresponding
to estimated values of 0 and 0 with the Gumbel-options MU=EST and SIGMA=EST. Specify these
options in parentheses following the GUMBEL option.
Agreement between the reference line and the point pattern indicates that the Gumbel distribution with
GRID
draws reference lines perpendicular to the quantile axis at major tick marks.
LEGEND=name | NONE
and fitted curves. Specifying LEGEND=NONE is equivalent to specifying the NOLEGEND option.
LOGNORMAL(SIGMA=value-list|EST < lognormal-options >)

LNORM(SIGMA=value-list|EST < lognormal-options >)
creates a lognormal Q-Q plot for each value of the shape parameter given by the mandatory SIGMA=
option or its alias, the SHAPE= option. For example,

qqplot width/ lognormal(shape=1.5 2.5);
run;

fromsmallest to largest, and the ith ordered observation
1 i 0:375
is plotted against the quantile exp ˆ nC0:25 , where ˆ 1 ./ is the inverse cumulative standard
normal distribution, n is the number of nonmissing observations, and is the shape parameter of the
lognormal distribution.
The pattern on the plot for SIGMA= tends to be linear with intercept and slope exp./ if the data
are lognormally distributed with the specific density function

.log.x / /2
(
p 1 exp 2 for x >
p.x/ D 2.x / 2
0 for x
where
D scale parameter
To obtain a graphical estimate of , specify a list of values for the SIGMA= option, and select the
value that most nearly linearizes the point pattern. For an illustration, see Example 6.24.
To assess the point pattern, you can add a diagonal distribution reference line corresponding to
the threshold parameter 0 and the scale parameter 0 with the lognormal-options THETA=0 and
ZETA=0 . Alternatively, you can add a line corresponding to estimated values of 0 and 0 with
the lognormal-options THETA=EST and ZETA=EST. This line has intercept 0 and slope exp.0 /.
Agreement between the line and the point pattern indicates that the lognormal distribution with
parameters , 0 , and 0 is a good fit. See Output 6.24.4 for an example. You can specify the
THRESHOLD= option as an alias for the THETA= option and the SCALE= option as an alias for the
ZETA= option.
You can also display the reference line by specifying THETA=0 , and you can specify the slope with
the SLOPE= option. For example, the following two QQPLOT statements produce charts with identical
reference lines:

qqplot width / lognormal(sigma=2 theta=3 zeta=1);
qqplot width / lognormal(sigma=2 theta=3 slope=2.718);
run;
MU=value|EST
specifies a value for the mean for a Q-Q plot requested with the GUMBEL and NORMAL options.
For the normal distribution, you can specify MU=EST to request a distribution reference line with
intercept equal to the sample mean, as illustrated in Figure 6.39. If you specify MU=EST for the
Gumbel distribution, a maximum likelihood estimate is calculated.
NADJ=value
specifies the adjustment value added to the sample size in the calculation of theoretical quantiles.
The default is 14 , as described by Blom (1958). Also refer to Chambers et al. (1983) for additional
information.
NOLEGEND
LEGEND=NONE
suppresses legends for specification limits, fitted curves, distribution lines, and hidden observations.
For an example, see Output 6.26.1.
NOLINELEGEND
NOLINEL
suppresses the legend for the optional distribution reference line.

NORM< (normal-options) >
creates a normal Q-Q plot. This is the default if you do not specify a distribution option. To create
the plot, the observations are ordered
from
smallest to largest, and the ith ordered observation is
1 i 0:375 1
plotted against the quantile ˆ nC0:25 , where ˆ ./ is the inverse cumulative standard normal
distribution, and n is the number of nonmissing observations.
The pattern on the plot tends to be linear with intercept and slope if the data are normally distributed
with the specific density function

p1 .x /2
p.x/ D exp 2 2
for all x
2
where is the mean, and is the standard deviation . > 0/.

To assess the point pattern, you can add a diagonal distribution reference line with intercept 0
and slope 0 with the normal-options MU=0 and SIGMA=0 . Alternatively, you can add a line
corresponding to estimated values of 0 and 0 with the normal-options MU=EST and SIGMA=EST;
the estimates of 0 and 0 are the sample mean and sample standard deviation. Specify these options
in parentheses, as in the following example:

qqplot length / normal(mu=10 sigma=0.3);
run;
For an example, see “Adding a Distribution Reference Line” on page 511. Agreement between the
reference line and the point pattern indicates that the normal distribution with parameters 0 and 0 is
a good fit. You can specify MU=EST and SIGMA=EST to request a distribution reference line with
the sample mean and sample standard deviation as the intercept and slope.
Other normal-options include CPKREF and CPKSCALE. The CPKREF option draws reference lines
extending from the intersections of specification limits with the distribution reference line to the
theoretical quantile axis. The CPKSCALE option rescales the theoretical quantile axis in Cpk units.
You can use the CPKREF option with the CPKSCALE option for graphical estimation of the capability
indices CPU , CPL, and Cpk , as illustrated in Output 6.26.1.
NOSPECLEGEND
NOSPECL
suppresses the legend for specification limit reference lines. For an example, see Figure 6.39.
PARETO(< Pareto-options >)

creates a generalized Pareto Q-Q plot for each value of the shape parameter ˛ given by the mandatory
estimate for ˛.
is plotted against the quantile .1 .1 inC0:250:375 ˛
/ /=˛ (˛ ¤ 0) or log.1 inC0:25 0:375
/ (˛ D 0), where
n is the number of nonmissing observations and ˛ is the shape parameter of the generalized Pareto
distribution.
are generalized Pareto distributed with the specific density function
.1 ˛.x /= /1=˛ 1 if ˛ ¤ 0

1
p.x/ D 1
exp. .x /= / if ˛ D 0
where
with the Pareto-options THETA=0 and SIGMA=0 . Alternatively, you can add a line corresponding
to estimated values of 0 and 0 with the Pareto-options THETA=EST and SIGMA=EST. Specify
these options in parentheses following the PARETO option.
Agreement between the reference line and the point pattern indicates that the generalized Pareto
distribution with parameters ˛, 0 , and 0 is a good fit.
PCTLAXIS(axis-options)
adds a nonlinear percentile axis along the frame of the Q-Q plot opposite the theoretical quantile axis.
The added axis is identical to the axis for probability plots produced with the PROBPLOT statement.
When using the PCTLAXIS option, you must specify HREF= values in quantile units, and you cannot
use the NOFRAME option. You can specify the following axis-options:
CGRID=color
specifies the color used for grid lines.
GRID
draws grid lines perpendicular to the percentile axis at major tick marks.
GRIDCHAR=‘character ’
specifies the character used to draw grid lines associated with the percentile axis on line printer
plots.
LABEL=‘string’
specifies the label for the percentile axis.
LGRID=linetype
specifies the line type used for grid lines associated with the percentile axis.
PCTLORDER=value-list
specifies the tick mark values labeled on the percentile axis. The values must be listed in
increasing order and must be between 0 and 100, exclusive. Values that correspond to quantiles
that are outside the range of the theoretical quantile axis are not displayed.
WGRID=value
specifies the thickness for grid lines associated with the percentile axis.
For example, the following statements display the plot in Figure 6.40:

qqplot Distance / normal(mu=est sigma=est)
nolegend
pctlaxis(grid label='Normal Percentiles')
odstitle=title;
run;
Figure 6.40 Normal Q-Q Plot with Percentile Axis
PCTLSCALE
requests scale labels for the theoretical quantile axis in percentile units, resulting in a nonlinear axis
scale. Tick marks are drawn uniformly across the axis based on the quantile scale. In all other respects,
the plot remains the same, and you must specify HREF= values in quantile units. For a true nonlinear
axis, use the PCTLAXIS option or use the PROBPLOT statement.
For example, the following statements display the plot in Figure 6.41:

qqplot Distance / normal(mu=est sigma=est cpkref)
pctlaxis(grid lgrid=35)
nolegend pctlscale
odstitle=title;
run;
Figure 6.41 Normal Q-Q Plot for Reading Percentiles of Specification Limits
POWER(< power-options >)

creates a power function Q-Q plot for each value of the shape parameter ˛ given by the mandatory
estimate for ˛.
To create the plot, the observations are orderedfrom smallest to largest, and the ith ordered observation
1 i 0:375 1
is plotted against the quantile B˛.1/ nC0:25 , where B˛.1/ ./ is the inverse normalized incomplete
beta function, n is the number of nonmissing observations, ˛ is one shape parameter of the beta
distribution, and the second shape parameter, ˇ D 1.
are power function distributed with the specific density function
( ˛ 1
˛ x
for < x < C
p.x/ D
0 for x or x C
where
with the power-options THETA=0 and SIGMA=0 . Alternatively, you can add a line corresponding
to estimated values of 0 and 0 with the power-options THETA=EST and SIGMA=EST. Specify
these options in parentheses following the POWER option.
Agreement between the reference line and the point pattern indicates that the power function distribution
with parameters ˛, 0 , and 0 is a good fit.
RANKADJ=value
specifies the adjustment value added to the ranks in the calculation of theoretical quantiles. The default
is 83 , as described by Blom (1958). Also refer to Chambers et al. (1983) for additional information.
RAYLEIGH(< Rayleigh-options >)

creates a Rayleigh Q-Q plot. To create the plot, the observations
r are ordered from smallest to largest,

and the ith ordered observation is plotted against the quantile 2 log 1 inC0:25
0:375
, where n is the
number of nonmissing observations.
The point pattern on the plot tends to be linear with intercept and slope if the data are Rayleigh
x
2
exp. .x /2 =.2 2 // for x
p.x/ D
0 for x <

with the Rayleigh-options THETA=0 and SIGMA=0 . Alternatively, you can add a line corresponding
to estimated values of 0 and 0 with the Rayleigh-options THETA=EST and SIGMA=EST. Specify
these options in parentheses after the RAYLEIGH option.
Agreement between the reference line and the point pattern indicates that the Rayleigh distribution
with parameters 0 and 0 is a good fit.
ROTATE
switches the horizontal and vertical axes so that the theoretical percentiles are plotted vertically while
the data are plotted horizontally. Regardless of whether the plot has been rotated, horizontal axis
options (such as HAXIS=) refer to the horizontal axis, and vertical axis options (such as VAXIS=)
refer to the vertical axis. All other options that depend on axis placement adjust to the rotated axes.
SIGMA=value-list|EST
specifies the value of the distribution parameter , where > 0. Alternatively, you can specify
SIGMA=EST to request an estimate for 0 that is computed as described in “Lognormal Distribution”
on page 353. The use of the SIGMA= option depends on the distribution option specified, as indicated
by Table 6.78.
Distribution Option Use of the SIGMA= Option

BETA THETA=0 and SIGMA=0 request a distribution reference
EXPONENTIAL line with intercept 0 and slope 0 .
GAMMA
PARETO
POWER
RAYLEIGH
WEIBULL
GUMBEL MU=0 and SIGMA=0 request a distribution reference line
corresponding to 0 and 0 .
LOGNORMAL SIGMA=1 : : : n requests n Q-Q plots with shape parameters
1 : : : n . The SIGMA= option is mandatory.
NORMAL MU=0 and SIGMA=0 request a distribution reference line
with intercept 0 and slope 0 . SIGMA=EST requests a slope
equal to the sample standard deviation.
WEIBULL2 SIGMA=0 and C=c0 request a distribution reference line with
intercept log.0 / and slope c10 .
For an example using SIGMA=EST, see Output 6.26.1. For an example of lognormal plots using the
SIGMA= option, see Example 6.24.
SLOPE=value|EST
specifies the slope for a distribution reference line requested with the LOGNORMAL and WEIBULL2
options.
When you use the SLOPE= option with the LOGNORMAL option, you must also specify a threshold
parameter value 0 with the THETA= option. Specifying the SLOPE= option is an alternative to
specifying ZETA=0 , which requests a slope of exp.0 /. See Output 6.24.4 for an example.
When you use the SLOPE= option with the WEIBULL2 option, you must also specify a scale parameter
value 0 with the SIGMA= option. Specifying the SLOPE= option is an alternative to specifying C=c0 ,
which requests a slope of c10 .
For example, the first and second QQPLOT statements that follow produce plots identical to those
produced by the third and fourth QQPLOT statements:

qqplot width / lognormal(sigma=2 theta=0 zeta=0);
qqplot width / weibull2(sigma=2 theta=0 c=0.25);
qqplot width / lognormal(sigma=2 theta=0 slope=1);
qqplot width / weibull2(sigma=2 theta=0 slope=4);
run;
For more information, see “Graphical Estimation” on page 535.
SQUARE
displays the Q-Q plot in a square frame. Compare Figure 6.38 with Figure 6.39. The default is a
rectangular frame.
THETA=value|EST
THRESHOLD=value|EST
specifies the lower threshold parameter for Q-Q plots requested with the BETA, EXPONENTIAL,
GAMMA, LOGNORMAL, PARETO, POWER, RAYLEIGH, WEIBULL, and WEIBULL2 options.
When used with the WEIBULL2 option, the THETA= option specifies the known lower threshold 0 ,
for which the default is 0. See Output 6.25.2 for an example.
When used with the other distribution options, the THETA= option specifies 0 for a distribution
reference line; alternatively in this situation, you can specify THETA=EST to request a maximum like-
lihood estimate for 0 . To request the line, you must also specify a scale parameter. See Output 6.24.4
for an example of the THETA= option with a lognormal Q-Q plot.
WEIBULL(C=value-list|EST < Weibull-options >)

WEIB(C=value-list < Weibull-options >)
creates a three-parameter Weibull Q-Q plot for each value of the shape parameter c given by the
mandatory C= option or its alias, the SHAPE= option. For example,

qqplot width / weibull(c=1.8 to 2.4 by 0.2);
run;
To create the plot, the observations are ordered from smallest to largest, and the ith ordered obser-
1c
vation is plotted against the quantile log 1 inC0:25
0:375
, where n is the number of nonmissing
observations, and c is the Weibull distribution shape parameter.
The pattern on the plot for C=c tends to be linear with intercept and slope if the data are Weibull
( c 1 c
c x x
exp for x >
p.x/ D
0 for x
where is the threshold parameter, is the scale parameter . > 0/, and c is the shape parameter (c >
0).
To obtain a graphical estimate of c, specify a list of values for the C= option, and select the value
that most nearly linearizes the point pattern. For an illustration, see Example 6.25. To assess the
point pattern, you can add a diagonal distribution reference line with intercept 0 and slope 0 with
the Weibull-options THETA=0 and SIGMA=0 . Alternatively, you can add a line corresponding to
estimated values of 0 and 0 with the Weibull-options THETA=EST and SIGMA=EST. Specify these
options in parentheses, as in the following example:

qqplot width / weibull(c=2 theta=3 sigma=4);
run;
Agreement between the reference line and the point pattern indicates that the Weibull distribution with
parameters c, 0 , and 0 is a good fit. You can specify the SCALE= option as an alias for the SIGMA=
WEIBULL2< (Weibull2-options) >
W2< (Weibull2-options) >
creates a two-parameter Weibull Q-Q plot. You should use the WEIBULL2 option when your data
have a known lower threshold 0 . You can specify the threshold value 0 with the THETA= option
or its alias, the THRESHOLD= option. If you are uncertain of the lower threshold value, you can
estimate 0 graphically by specifying a list of values for the THETA= option. Select the value that
most linearizes the point pattern. The default is 0 D 0.
To create the plot, the observations are ordered from smallest to largest, and the log of the shifted

ith ordered observation x.i / , log.x.i / 0 /, is plotted against the quantile log log 1 inC0:25
0:375
,
where n is the number of nonmissing observations. Unlike the three-parameter Weibull quantile, the
preceding expression is free of distribution parameters. This is why the C= shape parameter option is
not mandatory with the WEIBULL2 option.
1
The pattern on the plot for THETA=0 tends to be linear with intercept log. / and slope c if the data
are Weibull distributed with the specific density function
( c 1 c
c x 0 x 0
exp for x > 0
p.x/ D
0 for x 0
where 0 is a known lower threshold parameter, is a scale parameter . > 0/, and c is a shape
parameter (c > 0).
The advantage of a two-parameter Weibull plot over a three-parameter Weibull plot is that you can
visually estimate the shape parameter c and the scale parameter from the slope and intercept of
the point pattern; see Example 6.25 for an illustration of this method. The disadvantage is that the
two-parameter Weibull distribution applies only in situations where the threshold parameter is known.
See “Graphical Estimation” on page 535 for more information.
To assess the point pattern, you can add a diagonal distribution reference line corresponding to the scale
parameter 0 and shape parameter c0 with the Weibull2-options SIGMA=0 and C=c0 . Alternatively,
you can add a distribution reference line corresponding to estimated values of 0 and c0 with the
Weibull2-options SIGMA=EST and C=EST. This line has intercept log.0 / and slope c1 . Agreement
0
between the line and the point pattern indicates that the Weibull distribution with parameters c0 , 0 ,
and 0 is a good fit. You can specify the SCALE= option as an alias for the SIGMA= option and the
SHAPE= option as an alias for the C= option.
You can also display the reference line by specifying SIGMA=0 , and you can specify the slope with
the SLOPE= option. For example, the following QQPLOT statements produce identical plots:

qqplot width / weibull2(theta=3 sigma=4 c=2);
qqplot width / weibull2(theta=3 sigma=4 slope=0.5);
run;
ZETA=value|EST
specifies a value for the scale parameter for lognormal Q-Q plots requested with the LOGNORMAL
option. Specify THETA=0 and ZETA=0 to request a distribution reference line with intercept 0 and
slope exp.0 /.

CGRID=color
specifies the color for the grid lines associated with the quantile axis, requested by the GRID option.
LGRID=linetype
specifies the line type for the grid lines associated with the quantile axis, requested by the GRID option.
CPKREF
draws reference lines extending from the intersections of the specification limits with the distribution
reference line to the quantile axis in plots requested with the NORMAL option. Specify CPKREF in
parentheses after the NORMAL option. You can use the CPKREF option with the CPKSCALE option
for graphical estimation of the capability indices CPU , CPL, and Cpk , as illustrated in Output 6.26.1.
PCTLMINOR
requests minor tick marks for the percentile axis displayed when you use the PCTLAXIS option. See
the entry for the PCTLAXIS option for an example.
WGRID=n
specifies the width of the grid lines associated with the quantile axis, requested with the GRID option.
If you use the WGRID= option, you do not need to specify the GRID option.

NOOBSLEGEND
NOOBSL
suppresses the legend that indicates the number of hidden observations.
QQSYMBOL=‘character ’
specifies the character used to plot the Q-Q points in line printer plots. The default is the plus sign (C).
specifies the character used for a distribution reference line in a line printer plot. The default character
is the first letter of the distribution option keyword.
Details: QQPLOT Statement

construction of Q-Q plots
interpretation of Q-Q plots
distributions supported by the QQPLOT statement
graphical estimation of shape parameters, location and scale parameters, theoretical percentiles, and
capability indices
SYMBOL statement options
Construction of Quantile-Quantile and Probability Plots

Figure 6.42 illustrates how a Q-Q plot is constructed. First, the n nonmissing values of the variable are
ordered from smallest to largest:
x.1/ x.2/ x.n/
Then the ith orderedvalue x.i/ is represented on the plot by a point whose y-coordinate is x.i / and whose
i 0:375
x-coordinate is F 1 nC0:25 , where F ./ is the theoretical distribution with zero location parameter and
unit scale parameter.
Figure 6.42 Construction of a Q-Q Plot
You can modify the adjustment constants –0.375 and 0.25 with the RANKADJ= and NADJ= options. This
default combination is recommended by Blom (1958). For additional information, refer to Chambers et al.
(1983). Because x.i / is a quantile of the empirical cumulative distribution function (ecdf), a Q-Q plot
Details: QQPLOT Statement F 533
compares quantiles of the ecdf with quantiles of a theoretical distribution. Probability plots (see “PROBPLOT
Statement: CAPABILITY Procedure” on page 477) are constructed the same way, except that the x-axis is
scaled nonlinearly in percentiles.
Interpretation of Quantile-Quantile and Probability Plots

The following properties of Q-Q plots and probability plots make them useful diagnostics of how well a
specified theoretical distribution fits a set of measurements:
If the quantiles of the theoretical and data distributions agree, the plotted points fall on or near the line
y D x.
If the theoretical and data distributions differ only in their location or scale, the points on the plot fall
on or near the line y D ax C b. The slope a and intercept b are visual estimates of the scale and
location parameters of the theoretical distribution.
Q-Q plots are more convenient than probability plots for graphical estimation of the location and scale
parameters because the x-axis of a Q-Q plot is scaled linearly. On the other hand, probability plots are more
convenient for estimating percentiles or probabilities.
There are many reasons why the point pattern in a Q-Q plot may not be linear. Chambers et al. (1983) and
Fowlkes (1987) discuss the interpretations of commonly encountered departures from linearity, and these are
summarized in Table 6.79.
Table 6.79 Quantile-Quantile Plot Diagnostics
Description of Point Pattern Possible Interpretation

All but a few points fall on a line Outliers in the data
Left end of pattern is below the line; Long tails at both ends of the data
right end of pattern is above the line distribution
Left end of pattern is above the line; Short tails at both ends of the data
right end of pattern is below the line distribution
Curved pattern with slope increasing Data distribution is skewed to the right
from left to right
Curved pattern with slope decreasing Data distribution is skewed to the left
from left to right
Staircase pattern (plateaus and gaps) Data have been rounded or are discrete
In some applications, a nonlinear pattern may be more revealing than a linear pattern. However, Chambers
et al. (1983) note that departures from linearity can also be due to chance variation.

You can use the QQPLOT statement to request Q-Q plots based on the theoretical distributions summarized
in Table 6.80.
Table 6.80 QQPLOT Statement Distribution Options
Parameters
.x /˛ 1 .C x/ˇ 1
<x < C ˛, ˇ

1 x
Exponential exp x
˛ 1
1 x x
Gamma .˛/ exp x> ˛
.x /=

e
Gumbel exp e .x /= all x

p 1 .log.x / /2
Lognormal exp 2 2
x>
2.x /
(3-parameter)

p1 .x /2
Normal exp 2 2
all x
2
Generalized ˛¤0 1
.1 ˛.x /= /1=˛ 1
x> ˛
1
Pareto ˛D0 exp. .x /= /
˛ 1
˛ x
Power function x> ˛
x
Rayleigh 2
exp. .x /2 =.2 2 // x
c 1 c
c x x
Weibull exp x> c
(3-parameter)
c 1 c
c x 0 x 0
Weibull exp x > 0 0 c
(2-parameter) (known)
You can request these distributions with the BETA, EXPONENTIAL, GAMMA, LOGNORMAL, NORMAL,
WEIBULL, and WEIBULL2 options, respectively. If you do not specify a distribution option, a normal Q-Q
plot is created.
Graphical Estimation
You can use Q-Q plots to estimate shape, location, and scale parameters and to estimate percentiles. If you
are working with a normal Q-Q plot, you can also estimate certain capability indices.
Shape Parameters
Some distribution options in the QQPLOT statement require that you specify one or two shape parameters in
parentheses after the distribution keyword. These are summarized in Table 6.81.
You can visually estimate a shape parameter by specifying a list of values for the shape parameter option. A
separate plot is displayed for each value, and you can then select the value that linearizes the point pattern.
Alternatively, you can request that the plot be created using an estimated shape parameter. See the entries for
the distribution options in the section “Dictionary of Options” on page 517. for details on specification of
shape parameters. Example 6.24 and Example 6.25 illustrate shape parameter estimation with lognormal and
Weibull Q-Q plots.
Note that for Q-Q plots requested with the WEIBULL2 option, you can estimate the shape parameter c from
a linear pattern using the fact that the slope of the pattern is 1c . For an illustration, see Example 6.25.
Table 6.81 Shape Parameter Options for the

QQPLOT Statement
Distribution Keyword Mandatory Shape Parameter Option Range

BETA ALPHA=˛, BETA=ˇ ˛ > 0, ˇ > 0
EXPONENTIAL None
GAMMA ALPHA=˛ ˛>0
GUMBEL None
LOGNORMAL SIGMA= >0
NORMAL None
PARETO ALPHA=˛ ˛>0
POWER ALPHA=˛ ˛>0
RAYLEIGH None
WEIBULL C=c a>0
WEIBULL2 None
Location and Scale Parameters

When the point pattern on a Q-Q plot is linear, its intercept and slope provide estimates of the location and
scale parameters. (An exception to this rule is the two-parameter Weibull distribution, for which the intercept
and slope are related to the scale and shape parameters.) Table 6.82 shows how the intercept and slope are
related to the parameters for each distribution supported by the QQPLOT statement.
Table 6.82 Intercept and Slope of Linear Q-Q Plots
Parameters Linear Pattern

Distribution Location Scale Shape Intercept Slope
Beta ˛,ˇ
Exponential
Gamma ˛
Gumbel
Lognormal exp./
Normal
Generalized Pareto ˛
Power function ˛
Rayleigh
Weibull (3-parameter) c
1
Weibull (2-parameter) 0 (known) c log. / c
You can enhance a Q-Q plot with a diagonal distribution reference line by specifying the parameters that
determine the slope and intercept of the line; alternatively, you can request estimates for these parameters.
This line is an aid to checking the linearity of the point pattern, and it facilitates parameter estimation. For
instance, specifying MU=3 and SIGMA=2 with the NORMAL option requests a line with intercept 3 and
slope 2. Specifying SIGMA=1 and C=2 with the WEIBULL2 option requests a line with intercept log.1/ D 0
and slope 21 .
With the LOGNORMAL and WEIBULL2 options, you can specify the slope directly with the SLOPE=
option. That is, for the LOGNORMAL option, specifying THETA=0 and SLOPE=exp.0 / gives the same
reference line as specifying THETA=0 and ZETA=0 . For the WEIBULL2 option, specifying SIGMA=0
and SLOPE= c10 gives the same reference line as specifying SIGMA=0 and C=c0 .
For an example of parameter estimation using a normal Q-Q plot, see “Adding a Distribution Reference
Line” on page 511. Example 6.24 illustrates parameter estimation using a lognormal plot, and Example 6.25
illustrates estimation using two-parameter and three-parameter Weibull plots.
Theoretical Percentiles
There are two ways to estimate percentiles from a Q-Q plot:
Specify the PCTLAXIS option, which adds a percentile axis opposite the theoretical quantile axis. The
scale for the percentile axis ranges between 0 and 100 with tick marks at percentile values such as 1, 5,
10, 25, 50, 75, 90, 95, and 99. See Figure 6.40 for an example.
Specify the PCTLSCALE option, which relabels the horizontal axis tick marks with their percentile
equivalents but does not alter their spacing. For example, on a normal Q-Q plot, the tick mark labeled
“0” is relabeled as “50” because the 50th percentile corresponds to the zero quantile. See Figure 6.41
for an example.
You can also estimate percentiles using probability plots created with the PROBPLOT statement. See
Output 6.22.1 for an example.
Capability Indices
When the point pattern on a normal Q-Q plot is linear, you can estimate the capability indices CPU , CPL,
and Cpk from the plot, as explained by Rodriguez (1992). This method exploits the fact that the horizontal
axis of a Q-Q plot indicates the distance in standard deviation units (multiple of ) between a measurement
or specification limit and the process average.
In particular, one-third the standardized distance between an upper specification limit and the mean is the
one-sided capability index CPU .
USL
CPU D
3
Likewise, one-third the standardized distance between a lower specification limit and the mean is the
one-sided capability index CPL.
LSL
CPL D
3
Consequently, if you rescale the quantile axis of a normal Q-Q plot by a factor of three, you can read CPU
and CPL from the horizontal coordinates of the points at which the upper and lower specification lines
intersect the point pattern. Because Cpk is defined as the minimum of CPU and CPL this method also
provides a graphical estimate of Cpk . For an illustration, see Example 6.26.
SYMBOL Statement Options

In earlier releases of SAS/QC software, graphical features of lower and upper specification lines and diagonal
distribution reference lines were controlled with options in the SYMBOL2, SYMBOL3, and SYMBOL4
statements, respectively. These options are still supported, although they have been superseded by options in
the QQPLOT and SPEC statements. Table 6.83 summarizes the two sets of options. N OTE : These statements
have no effect on ODS Graphics output.
Table 6.83 SYMBOL Statement Options

Symbol markers SYMBOL1 statement
Character VALUE=special-symbol
Color COLOR=color
Font FONT=font
Height HEIGHT=value
Position LSL=value
Color CLSL=color COLOR=color
Line type LLSL=linetype LINE=linetype
Width WLSL=value WIDTH=value
Position USL=value
Color CUSL=color COLOR=color
Line type LUSL=linetype LINE=linetype
Width WUSL=value WIDTH=value

Target reference line SPEC statement

Color CTARGET=color
Line type LTARGET=linetype
Width WTARGET=value
Distribution reference line QQPLOT statement SYMBOL4 statement
Color COLOR=color COLOR=color
Line type LINE=linetype LINE=linetype
Width WIDTH=value WIDTH=value
ODS Graphics
destination where the graph is produced. QQPLOT options used to control the appearance of traditional
When ODS Graphics is in effect, the QQPLOT statement assigns a name to the graph it creates. You can use
this name to reference the graph when using ODS. The name is listed in Table 6.84.
Table 6.84 ODS Graphics Produced by the QQPLOT Statement

QQPlot Q-Q plot
producing charts.
Examples: QQPLOT Statement F 539
Examples: QQPLOT Statement

This section provides advanced examples of the QQPLOT statement.
Example 6.23: Interpreting a Normal Q-Q Plot of Nonnormal Data

N OTE : See Creating Lognormal Q-Q Plots in the SAS/QC Sample Library.
The following statements produce the normal Q-Q plot in Output 6.23.1:
data Measures;
input Diameter @@;
label Diameter='Diameter in mm';
datalines;
5.501 5.251 5.404 5.366 5.445 5.576 5.607
5.200 5.977 5.177 5.332 5.399 5.661 5.512
5.252 5.404 5.739 5.525 5.160 5.410 5.823
5.376 5.202 5.470 5.410 5.394 5.146 5.244
5.309 5.480 5.388 5.399 5.360 5.368 5.394
5.248 5.409 5.304 6.239 5.781 5.247 5.907
5.208 5.143 5.304 5.603 5.164 5.209 5.475
5.223
;
title 'Normal Q-Q Plot for Diameters';

proc capability data=Measures noprint;
qqplot Diameter / normal square odstitle=title;
run;
Output 6.23.1 Normal Quantile-Quantile Plot of Nonnormal Data
The nonlinearity of the points in Output 6.23.1 indicates a departure from normality. Because the point
pattern is curved with slope increasing from left to right, a theoretical distribution that is skewed to the right,
such as a lognormal distribution, should provide a better fit than the normal distribution. The mild curvature
suggests that you should examine the data with a series of lognormal Q-Q plots for small values of the shape
parameter, as illustrated in the next example.
Example 6.24: Estimating Parameters from Lognormal Plots

This example, which is a continuation of Example 6.23, demonstrates techniques for estimating the shape
parameter, location and scale parameters, and theoretical percentiles for a lognormal distribution.
Three-Parameter Lognormal Plots

The three-parameter lognormal distribution depends on a threshold parameter , a scale parameter , and
a shape parameter . You can estimate from a series of lognormal Q-Q plots with different values of .
The estimate is the value of that linearizes the point pattern. You can then estimate the threshold and scale
parameters from the intercept and slope of the point pattern. The following statements create the series of
plots in Output 6.24.1 through Output 6.24.3 for values of 0.2, 0.5, and 0.8:
title 'Lognormal Q-Q Plot for Diameters';

qqplot Diameter / lognormal(sigma=0.2 0.5 0.8)
square
odstitle=title;
run;
Output 6.24.1 Lognormal Quantile-Quantile Plot ( =0.2)


N OTE : You must specify a value for the shape parameter for a lognormal Q-Q plot with the SIGMA=
option or its alias, the SHAPE= option.
The plot in Output 6.24.2 displays the most linear point pattern, indicating that the lognormal distribution
with D 0:5 provides a reasonable fit for the data distribution.
Data with this particular lognormal distribution have the density function
( p
2
exp 2.log.x / /2

p for x >
p.x/ D .x /
0 for x
The points in the plot fall on or near the line with intercept and slope exp./. Based on Output 6.24.2,
5 and exp./ 1:2 3 D 0:4, giving log.0:4/ 0:92.
Estimating Percentiles
You can use a Q-Q plot to estimate percentiles such as the 95th percentile of the lognormal distribution.7
The point pattern in Output 6.24.2 has a slope of approximately 0.39 and an intercept of 5. The following
statements reproduce this plot, adding a lognormal reference line with this slope and intercept.
title 'Lognormal Q-Q Plot for Diameters';

qqplot Diameter / lognormal(sigma=0.5 theta=5 slope=0.39)
pctlaxis(grid)
vref = 5.8 5.9 6.0;
run;
Graphics, so the Q-Q plot is created using ODS Graphics instead of traditional graphics. The result is shown
in Output 6.24.4.
7 You can also use a probability plot for this purpose. See Output 6.22.1.
Output 6.24.4 Lognormal Q-Q Plot Identifying Percentiles
The PCTLAXIS option labels the major percentiles, and the GRID option draws percentile axis reference
lines. The 95th percentile is 5.9, because the intersection of the distribution reference line and the 95th
reference line occurs at this value on the vertical axis.
Alternatively, you can compute this percentile from the estimated lognormal parameters. The 100˛th
percentile of the lognormal distribution is
1
P˛ D exp. ˆ .˛/ C / C
where ˆ 1 ./
is the inverse cumulative standard normal distribution. Consequently,

1 1 1
P0:95 exp ˆ .0:95/ C log.0:39/ C 5 exp 1:645 0:94 C 5 5:89
2 2
Two-Parameter Lognormal Plots

If a known threshold parameter is available, you can construct a two-parameter lognormal Q-Q plot by
subtracting the threshold from the data and requesting a normal Q-Q plot. The following statements create
this plot for Diameter, assuming a known threshold of five:
data Measures;
set Measures;
Logdiam=log(Diameter-5);
label Logdiam='log(Diameter-5)';
run;
title 'Two-Parameter Lognormal Q-Q Plot for Diameters';

qqplot Logdiam / normal(mu=est sigma=est)
square
odstitle=title;
run;
Output 6.24.5 Two-Parameter Lognormal Q-Q Plot for Diameters
Because the point pattern in Output 6.24.5 is linear, you can estimate the lognormal parameters and as
the normal plot estimates of and , which are –0.99 and 0.51. These values correspond to the previous
estimates of –0.92 for and 0.5 for .
Example 6.25: Comparing Weibull Q-Q Plots

N OTE : See Creating Weibull Q-Q Plots in the SAS/QC Sample Library.
This example compares the use of three-parameter and two-parameter Weibull Q-Q plots for the failure times
in months for 48 integrated circuits. The times are assumed to follow a Weibull distribution.
data Failures;
input Time @@;
label Time='Time in Months';
datalines;
29.42 32.14 30.58 27.50 26.08 29.06 25.10 31.34
29.14 33.96 30.64 27.32 29.86 26.28 29.68 33.76
29.32 30.82 27.26 27.92 30.92 24.64 32.90 35.46
30.28 28.36 25.86 31.36 25.26 36.32 28.58 28.88
26.72 27.42 29.02 27.54 31.60 33.46 26.78 27.82
29.18 27.94 27.66 26.42 31.00 26.64 31.44 32.52
;
Three-Parameter Weibull Plots

If no assumption is made about the parameters of this distribution, you can use the WEIBULL option to
request a three-parameter Weibull plot. As in the previous example, you can visually estimate the shape
parameter c by requesting plots for different values of c and choosing the value of c that linearizes the point
pattern. Alternatively, you can request a maximum likelihood estimate for c, as illustrated in the following
statements produce Weibull plots for c = 1, 2 and 3:
title 'Three-Parameter Weibull Q-Q Plot for Failure Times';

proc capability data=Failures noprint;
qqplot Time / weibull(c=est theta=est sigma=est)
square
href=0.5 1 1.5 2
vref=25 27.5 30 32.5 35
odstitle=title;
run;
N OTE : When using the WEIBULL option, you must either specify a list of values for the Weibull shape
parameter c with the C= option, or you must specify C=EST.
Output 6.25.1 displays the plot for the estimated value c = 1.99. The reference line corresponds to the
estimated values for the threshold and scale parameters of (O0 =24.19 and O0 =5.83, respectively.
Output 6.25.1 Three-Parameter Weibull Q-Q Plot for c = 2
Two-Parameter Weibull Plots

N OTE : See Creating Weibull Q-Q Plots in the SAS/QC Sample Library.
Now, suppose it is known that the circuit lifetime is at least 24 months. The following statements use the
threshold value 0 D 24 to produce the two-parameter Weibull Q-Q plot shown in Output 6.25.2:
title 'Two-Parameter Weibull Q-Q Plot for Failure Times';

proc capability data=Failures noprint;
qqplot Time / weibull2(theta=24 c=est sigma=est) square
href= -4 to 1
vref= 0 to 2.5 by 0.5
odstitle=title;
run;
O
The reference line is based on maximum likelihood estimates c=2.08 O
and =6.05. These estimates agree
with those of the previous example.
Output 6.25.2 Two-Parameter Weibull Q-Q Plot for 0 D 24
Example 6.26: Estimating Cpk from a Normal Q-Q Plot

This example illustrates how you can use a normal Q-Q plot to estimate the capability index Cpk . The data
used here are the distance measurements provided in the section “Creating a Normal Quantile-Quantile Plot”
on page 509.
The linearity of the point pattern in Figure 6.39 indicates that the measurements are normally distributed (recall
that normality should be checked when process capability indices are reported). Furthermore, Figure 6.39
shows that the upper specification limit is about 1.7 standard deviation units above the mean, and the lower
specification limit is about 1.8 standard deviation units below the mean. Because CPU is defined as
USL
CPU D
3
and CPL is defined as
LSL
CPL D
3
it follows that an estimate of CPU is 1:7=3 D 0:57, and an estimate of CPL is 1:8=3 D 0:6. Thus, except for
a factor of three, you can estimate CPU and CPL from the points of intersection between the specification
lines and the point pattern.
The following statements facilitate this type of estimation by creating a Q-Q plot, displayed in Output 6.26.1,
in which the horizontal axis is rescaled by a factor of three:
ods graphics off;

symbol v=plus;
title "Normal Quantile-Quantile Plot for Hole Distance";
spec lsl=9.5 llsl=2 clsl=blue
usl=10.5 lusl=20 cusl=blue;
qqplot Distance / normal(mu=est sigma=est cpkscale cpkref)
nolegend
square;
inset n mean (5.2) std="Std Dev" (4.2) cpk (4.2) /
pos=(-0.75,10.48) data refpoint=tl;
run;
The CPKSCALE option rescales the horizontal axis, and the CPKREF option adds reference lines indicating
the intersections of the distribution reference line and the specification limits.
Output 6.26.1 Normal Q-Q Plot With Cpk Scaling

Using this display, you can estimate CPU and CPL directly from the horizontal axis as 0.55 and 0.60,
respectively (the negative sign for –0.60 is ignored). The minimum of these values (0.55) is an estimate of
Cpk . Note that this estimate agrees with the numerically obtained estimate for Cpk that is displayed on the
plot with the INSET statement.
See Rodriguez (1992) for further discussion concerning the use of Q-Q plots in process capability analysis.
Dictionary of Common Options: CAPABILITY Procedure

This chapter provides detailed descriptions of options that you can specify in the following chart statements:
CDFPLOT
COMPHISTOGRAM
HISTOGRAM
PPPLOT
PROBPLOT
QQPLOT
As noted, some options are applicable only to comparative plots produced by the COMPHISTOGRAM
statement or by another plot statement in conjunction with a CLASS statement.
General Options
ALPHADELTA=value
specifies the change in successive estimates of ˛O at which iteration terminates in the Newton-Raphson
approximation of the maximum likelihood estimate of ˛ for gamma distributions requested with the
GAMMA option. Enclose the ALPHADELTA= option in parentheses after the GAMMA keyword.
Iteration continues until the change in ˛ is less than the value specified or the number of iterations
exceeds the value of the MAXITER= option. The default value is 0.00001.
ALPHAINITIAL=value
specifies the initial value for ˛O in the Newton-Raphson approximation of the maximum likelihood
estimate of ˛ for gamma distributions requested with the GAMMA option. Enclose the ALPHAINI-
TIAL= option in parentheses after the GAMMA keyword. The default value is Thom’s approximation
of the estimate of ˛. See Johnson, Kotz, and Balakrishnan (1995).
CDELTA=value
specifies the change in successive estimates of c at which iterations terminate in the Newton-Raphson
approximation of the maximum likelihood estimate of c for Weibull distributions requested by the
WEIBULL option. Enclose the CDELTA= option in parentheses after the WEIBULL keyword.
Iteration continues until the change in c between consecutive steps is less than the value specified
or until the number of iterations exceeds the value of the MAXITER= option. The default value is
0.00001.
General Options F 551
CINITIAL=value
specifies the initial value for cO in the Newton-Raphson approximation of the maximum likelihood
estimate of c for Weibull distributions requested with the WEIBULL or WEIBULL2 option. The
default value is 1.8. See Johnson, Kotz, and Balakrishnan (1995).
CONTENTS=‘string’
specifies the table of contents grouping entry for output produced by the plot statement. You can
specify CONTENTS=‘’ to suppress the grouping entry.
CPROP
CPROP=color | EMPTY
specifies the color for a horizontal bar whose length (relative to the width of the tile) indicates the
proportion of the total frequency that is represented by the corresponding cell in a comparative plot. By
default, no proportion bars are displayed. You can specify the keyword EMPTY to display empty bars.
For traditional graphics with the GSTYLE system option in effect, you can specify CPROP with no
argument to produce proportion bars using an appropriate color from the ODS style. The CPROP
option is not available with ODS Graphics.
HAXIS=value
specifies the name of an AXIS statement describing the horizontal axis.
HREF=values
draws reference lines that are perpendicular to the horizontal axis at the values that you specify. Also
see the CHREF= and LHREF= options.
HREFLABELS=‘label1’ . . . ‘labeln’
HREFLABEL=‘label1’ . . . ‘labeln’
HREFLAB=‘label1’ . . . ‘labeln’
specifies labels for the lines requested by the HREF= option. The number of labels must equal the
number of lines. Enclose each label in quotes. Labels can have up to 16 characters.
HREFLABPOS=n
specifies the vertical position of HREFLABELS= labels, as described in the following table:
n Position
1 Along top of plot
2 Staggered from top to bottom of plot
3 Along bottom of plot
4 Staggered from bottom to top of plot
By default, HREFLABPOS=1. N OTE : HREFLABPOS=2 and HREFLABPOS=4 are not supported

for ODS Graphics output.
INTERTILE=value
specifies the distance in horizontal percentage screen units between the framed areas, called tiles,
of a comparative plot. By default, INTERTILE=0.75 percentage screen units. You can specify
INTERTILE=0 to create contiguous tiles.
MAXITER=n
specifies the maximum number of iterations in the Newton-Raphson approximation of the maximum
likelihood estimate of ˛ for gamma distributions requested with the GAMMA option and c for Weibull
distributions requested with the WEIBULL and WEIBULL2 options. Enclose the MAXITER= option
in parentheses after the GAMMA, WEIBULL, or WEIBULL2 keywords. The default value of n is 20.
NCOLS=n
NCOL=n
specifies the number of columns per panel in a comparative plot. By default, NCOLS=1 if you specify
only one CLASS variable, and NCOLS=2 if you specify two CLASS variables. If you specify two
CLASS variables, you can use the NCOLS= option with the NROWS= option.
NOHLABEL
suppresses the label for the horizontal axis. You can use this option to reduce clutter.
NOVLABEL
suppresses the label for the vertical axis. You can use this option to reduce clutter.
NOVTICK
suppresses the tick marks and tick mark labels for the vertical axis. This option also suppresses the
label for the vertical axis.
NROWS=n
NROW=n
specifies the number of rows per panel in a comparative plot. By default, NROWS=2. If you specify
two CLASS variables, you can use the NCOLS= option with the NROWS= option.
ODSFOOTNOTE=FOOTNOTE | FOOTNOTE1 | ‘string’

adds a footnote to ODS Graphics output. If you specify the FOOTNOTE (or FOOTNOTE1) keyword,
the value of SAS FOOTNOTE statement is used as the graph footnote. If you specify a quoted string,
that is used as the footnote. The quoted string can contain either of the following escaped characters,
which are replaced with the appropriate values from the analysis:
nn analysis variable name

nl analysis variable label (or name if the analysis variable has no label)
ODSFOOTNOTE2=FOOTNOTE2 | ‘string’
adds a secondary footnote to ODS Graphics output. If you specify the FOOTNOTE2 keyword, the
value of SAS FOOTNOTE2 statement is used as the secondary graph footnote. If you specify a quoted
string, that is used as the secondary footnote. The quoted string can contain any of the following
escaped characters, which are replaced with the appropriate values from the analysis:

General Options F 553
ODSTITLE=TITLE | TITLE1 | NONE | DEFAULT | LABELFMT | ‘string’

specifies a title for ODS Graphics output.
TITLE (or TITLE1) uses the value of SAS TITLE statement as the graph title.
NONE suppresses all titles from the graph.
DEFAULT uses the default ODS Graphics title (a descriptive title consisting of the plot type
and the analysis variable name.)
LABELFMT uses the default ODS Graphics title with the variable label instead of the variable
name.
If you specify a quoted string, that is used as the graph title. The quoted string can contain the following
escaped characters, which are replaced with the appropriate values from the analysis:

ODSTITLE2=TITLE2 | ‘string’
specifies a secondary title for ODS Graphics output. If you specify the TITLE2 keyword, the value
of SAS TITLE2 statement is used as the secondary graph title. If you specify a quoted string, that is
used as the secondary title. The quoted string can contain the following escaped characters, which are
replaced with the appropriate values from the analysis:

OVERLAY
specifies that plots associated with different levels of a CLASS variable be overlaid onto a single plot,
rather than displayed as separate cells in a comparative plot. If you specify the OVERLAY option with
one CLASS variable, the output associated with each level of the CLASS variable is overlaid on a
single plot. If you specify the OVERLAY option with two CLASS variables, a comparative plot based
on the first CLASS variable’s levels is produced. Each cell in this comparative plot contains overlaid
output associated with the levels of the second CLASS variable.
The OVERLAY option applies only to ODS Graphics output. It is not available in the COMPHIS-
TOGRAM statement.
SCALE=value
is an alias for the SIGMA= option for distributions requested by the BETA, EXPONENTIAL, GAMMA,
SB, SU, WEIBULL, and WEIBULL2 options and for the ZETA= option for distributions requested by
the LOGNORMAL option.
SHAPE=value
is an alias for the ALPHA= option for distributions requested by the GAMMA option, for the SIGMA=
option for distributions requested by the LOGNORMAL option, and for the C= option for distributions
requested by the WEIBULL and WEIBULL2 options.
STATREF=keyword-list
draws reference lines at the values of the statistics requested in the keyword-list . These reference lines
are perpendicular to the horizontal axis in a histogram or cdf plot, and perpendicular to the vertical
axis in a probability or Q-Q plot (unless the ROTATE option is specified). The STATREF= option does
not apply to the PPPLOT statement.
Valid keywords are listed in the following table:
Keyword Statistic
MAX Largest value
MEAN Sample mean
MEDIAN | Q2 Median (50th percentile)
MIN Smallest value
MODE Most frequent value
P pctl pctlth percentile
Q1 Lower quartile (25th percentile)
Q3 Upper quartile (75th percentile)
factor STD factor standard deviations from the mean
Note that the factor specified with the STD keyword can be positive (which puts a reference line above
the mean) or negative (below the mean).
Also see the CSTATREF=, LSTATREF=, STATREFLABELS=, and STATREFSUBCHAR= options.
STATREFLABELS=‘label1’ . . . ‘labeln’
STATREFLABEL=‘label1’ . . . ‘labeln’
STATREFLAB=‘label1’ . . . ‘labeln’
specifies labels for the lines requested by the STATREF= option. The number of labels must equal the
STATREFSUBCHAR=‘keyword-list’
specifies a substitution character (such as #) for labels specified with the STATREFLABELS= option.
When the labels are displayed on a graph, the first occurrence of the specified character in each label is
replaced with the value of the corresponding STATREF= statistic.
VAXIS=name
VAXIS=value-list
specifies the name of an AXIS statement describing the vertical axis. In a COMPHISTOGRAM or
HISTOGRAM statement, you can alternatively specify a value-list for the vertical axis.
VAXISLABEL=‘label’
specifies a label for the vertical axis. Labels can have up to 40 characters.
VREF=value-list
draws reference lines perpendicular to the vertical axis at the values specified. Also see the CVREF=
and LVREF= options.
Options for Traditional Graphics F 555
VREFLABELS=‘label1’. . . ‘labeln’
VREFLABEL=‘label1’. . . ‘labeln’
VREFLAB=‘label1’. . . ‘labeln’
specifies labels for the lines requested by the VREF= option. The number of labels must equal the
VREFLABPOS=n
specifies the horizontal position of VREFLABELS= labels. If you specify VREFLABPOS=1, the
labels are positioned at the left of the plot. If you specify VREFLABPOS=2, the labels are positioned
at the right of the plot. By default, VREFLABPOS=1 for traditional graphics and 2 for ODS Graphics.

ANNOKEY
applies the annotation requested with the ANNOTATE= option only to the key cell of a comparative
plot. By default, the procedure applies annotation to all of the cells. You can use the KEYLEVEL=
option in the CLASS statement or the CLASSKEY= option in the COMPHISTOGRAM statement to
specify the key cell.
ANNOTATE=SAS-data-set
ANNO=SAS-data-set
specifies an input data set that contains annotate variables, as described in SAS/GRAPH: Reference, for
annotating traditional graphics. The ANNOTATE= data set you specify in the plot statement is used
for all plots created by the statement. You can also specify an ANNOTATE= data set in the PROC
CAPABILITY statement to enhance all plots created by the procedure (see “ANNOTATE= Data Sets”
on page 232).
CAXIS=color
CAXES=color
CA=color
specifies the color for the axes and tick marks. This option overrides any COLOR= specifications in an
AXIS statement.
CFRAME=color
specifies the color for the area that is enclosed by the axes and frame.
CFRAMESIDE=color
specifies the color to fill the frame area for the row labels that display along the left side of a comparative
plot. This color also fills the frame area for the label of the corresponding CLASS variable, if you
associate a label with the variable.
CFRAMETOP=color
specifies the color to fill the frame area for the column labels that display across the top of a comparative
plot. This color also fills the frame area for the label of the corresponding CLASS variable, if you
associate a label with the variable.
CHREF=color | (color-list)
CH=color | (color-list)
specifies the colors for horizontal axis reference lines requested by the HREF= option. If you specify a
single color, it is used for all HREF= lines. Otherwise, if there are fewer colors specified than reference
lines requested, the remaining lines are displayed with the default reference line color. You can also
specify the value _default in the color list to request the default color.
COLOR=color
COLOR=color-list
specifies the color of the curve or reference line associated with a distribution or kernel density estimate.
Enclose the COLOR= option in parentheses after a distribution option or the KERNEL option. In a
HISTOGRAM statement, you can specify a list of colors in parentheses for multiple density curves.
CSTATREF=color | (color-list)
specifies the colors for reference lines requested by the STATREF= option. If you specify a single
color, it is used for all STATREF= lines. Otherwise, if there are fewer colors specified than reference
lines requested, the remaining lines are displayed with the default reference line color. You can also
specify the value _default in the color list to request the default color.
CTEXT=color
CT=color
specifies the color for tick mark values and axis labels.
CTEXTSIDE=color
specifies the color for the row labels that display along the left side of a comparative plot. If you do
not specify the CTEXTSIDE= option, the color specified with the CTEXT= option is used. You can
specify the CFRAMESIDE= option to change the background color for the row labels.
CTEXTTOP=color
specifies the color for the column labels that display along the left side of a comparative plot. If you do
not specify the CTEXTTOP= option, the color specified with the CTEXT= option is used. You can use
the CFRAMETOP= option to change the background color for the column labels.
CVREF=color | (color-list)
CV=color | (color-list)
specifies the colors for lines requested with the VREF= option. If you specify a single color, it is used
for all VREF= lines. Otherwise, if there are fewer colors specified than reference lines requested, the
remaining lines are displayed with the default reference line color. You can also specify the value
_default in the color list to request the default color.
DESCRIPTION=‘string’
DES=‘string’
specifies a description, up to 256 characters long, for the GRSEG catalog entry for a traditional graphics
chart. The default value is the analysis variable name.
Options for Traditional Graphics F 557
FONT=font
specifies a font for reference line and axis labels. You can also specify fonts for axis labels in an AXIS
statement. The FONT= option takes precedence over the FTEXT= font specified in the GOPTIONS
statement. For a list of software fonts, see SAS/GRAPH: Reference.
HEIGHT=value
specifies the height, in percentage screen units, of text for axis labels, tick mark labels, and legends.
This option takes precedence over the HTEXT= option in the GOPTIONS statement.
HMINOR=n
HM=n
specifies the number of minor tick marks between each major tick mark on the horizontal axis. Minor
tick marks are not labeled. By default, HMINOR=0.
INFONT=font
specifies a font to use for text inside the framed areas of the plot. The INFONT= option takes
precedence over the FTEXT= option in the GOPTIONS statement. For a list of software fonts, see
SAS/GRAPH: Reference.
INHEIGHT=value
specifies the height, in percentage screen units, of text used inside the framed areas of the plot. If you
do not specify the INHEIGHT= option, the height specified with the HEIGHT= option is used.
L=linetype
L=linetype-list
specifies the line type of the curve or reference line associated with a distribution or kernel density
estimate. Enclose the L= option in parentheses after the distribution option or the KERNEL option.
In a HISTOGRAM statement, you can specify a list of line types in parentheses for multiple density
curves.
LHREF=linetype | linetype-list
LH=linetype | linetype-list
specifies the line types for the reference lines that you request with the HREF= option. If you specify a
single line type, it is used for all HREF= lines. Otherwise, if there are fewer line types specified than
reference lines requested, the remaining lines are displayed with the default reference line type. You
can also specify line type 0 to request the default color.
LSTATREF=linetype | linetype-list
specifies the line types for the reference lines that you request with the STATREF= option. If you
specify a single line type, it is used for all STATREF= lines. Otherwise, if there are fewer line types
specified than reference lines requested, the remaining lines are displayed with the default reference
line type. You can also specify line type 0 to request the default color.
LVREF=linetype | linetype-list
LV=linetype | linetype-list
specifies the line types for lines requested with the VREF= option. If you specify a single line type,
it is used for all VREF= lines. Otherwise, if there are fewer line types specified than reference lines
requested, the remaining lines are displayed with the default reference line type. You can also specify
line type 0 to request the default color.
NAME=‘string’
specifies the name of the GRSEG catalog entry for a traditional graphics plot, and the name of the
graphics output file if one is created. The name can be up to 256 characters long, but the GRSEG name
is truncated to eight characters. The default value is ‘CAPABILI’.
NOFRAME
suppresses the frame around the subplot area.
TURNVLABELS
TURNVLABEL
turns the characters in the vertical axis labels so that they display vertically.
VMINOR=n
VM=n
specifies the number of minor tick marks between each major tick mark on the vertical axis. Minor
tick marks are not labeled. The default is zero.
W=value
W=value-list
specifies the width in pixels of the curve or reference line associated with a distribution or kernel
density estimate. Enclose the W= option in parentheses after the distribution option or the KERNEL
option. In a HISTOGRAM statement, you can specify a list of widths in parentheses for multiple
density curves.
WAXIS=n
specifies the line thickness, in pixels, for the axes and frame.
Options for Legacy Line Printer Charts

HREFCHAR=‘character ’
specifies the character used to form the lines requested by the HREF= option for a line printer chart.
The default is the vertical bar (|).
VREFCHAR=‘character ’
VREF= option for a line printer chart. specifies the character used to form the lines requested by the
The default is the hyphen (-).
References
Bai, D. S., and Choi, I. S. (1997). “Process Capability Indices for Skewed Populations.” Unpublished
manuscript, Korean Advanced Institute of Science and Technology, Taejon, Korea.
Bissell, A. F. (1990). “How Reliable Is Your Capability Index?” Journal of the Royal Statistical Society,
Series C 39:331–340.
Blom, G. (1958). Statistical Estimates and Transformed Beta Variables. New York: John Wiley & Sons.
References F 559
Bowman, K. O., and Shenton, L. R. (1983). “Johnson’s System of Distributions.” In Encyclopedia of

Statistical Sciences, vol. 4, edited by S. Kotz, N. L. Johnson, and C. B. Read. New York: John Wiley &
Sons.
Boyles, R. A. (1991). “The Taguchi Capability Index.” Journal of Quality Technology 23:107–126.
Boyles, R. A. (1992). Cpm for Asymmetrical Tolerances. Technical report, Precision Castparts Corp.,
Portland, OR.
Boyles, R. A. (1994). “Process Capability with Asymmetric Tolerances.” Communications in Statistics—

Simulation and Computation 23:615–643.
Chambers, J. M., Cleveland, W. S., Kleiner, B., and Tukey, P. A. (1983). Graphical Methods for Data
Analysis. Belmont, CA: Wadsworth International Group.
Chen, H. F., and Kotz, S. (1996). “An Asymptotic Distribution of Wright’s Process Capability Index Sensitive
to Skewness.” Journal of Statistical Computation and Simulation 55:147–158.
Chen, K. S. (1998). “Incapability Index with Asymmetric Tolerances.” Statistica Sinica 8:253–262.
Chou, Y., Owen, D. B., and Borrego, S. A. (1990). “Lower Confidence Limits on Process Capability Indices.”
Journal of Quality Technology 22:223–229; corrigenda, 24, 251.
Cohen, A. C. (1951). “Estimating Parameters of Logarithmic-Normal Distributions by Maximum Likelihood.”

Journal of the American Statistical Association 46:206–212.
Conover, W. J. (1980). Practical Nonparametric Statistics. 2nd ed. New York: John Wiley & Sons.
Croux, C., and Rousseeuw, P. J. (1992). “Time-Efficient Algorithms for Two Highly Robust Estimators of
Scale.” Computational Statistics 1:411–428.
D’Agostino, R. B., and Stephens, M., eds. (1986). Goodness-of-Fit Techniques. New York: Marcel Dekker.
Dixon, W. J., and Tukey, J. W. (1968). “Approximate Behavior of the Distribution of Winsorized t (Trim-
ming/Winsorization 2).” Technometrics 10:83–98.
Ekvall, D. N., and Juran, J. M. (1974). “Manufacturing Planning.” In Quality Control Handbook, 3rd ed.,
edited by J. M. Juran. New York: McGraw-Hill.
Elandt, R. C. (1961). “The Folded Normal Distribution: Two Methods of Estimating Parameters from
Moments.” Technometrics 3:551–562.
Fowlkes, E. B. (1987). A Folio of Distributions: A Collection of Theoretical Quantile-Quantile Plots. New

York: Marcel Dekker.
Gnanadesikan, R. (1997). Statistical Data Analysis of Multivariate Observations. New York: John Wiley &
Sons.
Grimshaw, S. D. (1993). “Computing Maximum Likelihood Estimates for the Generalized Pareto Distribution.”
Technometrics 35:185–191.
Gupta, A. K., and Kotz, S. (1997). “A New Process Capability Index.” Metrika 45:213–224.
Hahn, G. J. (1969). “Factors for Calculating Two-Sided Prediction Intervals for Samples from a Normal
Distribution.” Journal of the American Statistical Association 64:878–898.
Hahn, G. J. (1970a). “Additional Factors for Calculating Prediction Intervals for Samples from a Normal
Distribution.” Journal of the American Statistical Association 65:1668–1676.
Hahn, G. J. (1970b). “Statistical Intervals for a Normal Population, Part 2: Formulas, Assumptions, Some
Derivations.” Journal of Quality Technology 2:195–206.
Hahn, G. J., and Meeker, W. Q. (1991). Statistical Intervals: A Guide for Practitioners. New York: John
Wiley & Sons.
Iman, R. L. (1974). “Use of a t-Statistic as an Approximation to the Exact Distribution of the Wilcoxon
Signed Rank Statistic.” Communications in Statistics 3:795–806.
Johnson, N. L., Kotz, S., and Balakrishnan, N. (1994). Continuous Univariate Distributions. 2nd ed. Vol. 1.
New York: John Wiley & Sons.
Johnson, N. L., Kotz, S., and Balakrishnan, N. (1995). Continuous Univariate Distributions. 2nd ed. Vol. 2.
New York: John Wiley & Sons.
Johnson, N. L., Kotz, S., and Kemp, A. W. (1992). Univariate Discrete Distributions. 2nd ed. New York:
John Wiley & Sons.
Johnson, N. L., Kotz, S., and Pearn, W. L. (1994). “Flexible Process Capability Indices.” Pakistan Journal of
Statistics 10:23–31.
Kane, V. E. (1986). “Process Capability Indices.” Journal of Quality Technology 1:41–52.
Kotz, S., and Johnson, N. L. (1993). Process Capability Indices. London: Chapman & Hall.
Kotz, S., and Lovelace, C. R. (1998). Process Capability Indices in Theory and Practice. London: Edward
Arnold.
Krishnamoorthy, K., and Mathew, T. (2009). Statistical Tolerance Regions: Theory, Applications, and
Computation. Hoboken, NJ: John Wiley & Sons.
Kushler, R. H., and Hurley, P. (1992). “Confidence Bounds for Capability Indices.” Journal of Quality
Technology 24:188–195.
Lehmann, E. L., and D’Abrera, H. J. M. (1975). Nonparametrics: Statistical Methods Based on Ranks. San
Francisco: Holden-Day.
Luceño, A. (1996). “A Process Capability Index with Reliable Confidence Intervals.” Communications in
Statistics—Simulation and Computation 25:235–245.
Marcucci, M. O., and Beazley, C. F. (1988). “Capability Indices: Process Performance Measures.” Transac-
tions of ASQC Congress 42:516–523.
Montgomery, D. C. (1996). Introduction to Statistical Quality Control. 3rd ed. New York: John Wiley &
Sons.
Odeh, R. E., and Owen, D. B. (1980). Tables for Normal Tolerance Limits, Sampling Plans, and Screening.
New York: Marcel Dekker.
Owen, D. B., and Hua, T. A. (1977). “Tables of Confidence Limits on the Tail Area of the Normal Distribution.”
Communications in Statistics—Simulation and Computation 6:285–311.
References F 561
Pearn, W. L., Kotz, S., and Johnson, N. L. (1992). “Distributional and Inferential Properties of Process
Capability Indices.” Journal of Quality Technology 24:216–231.
Rodriguez, R. N. (1992). “Recent Developments in Process Capability Analysis.” Journal of Quality

Technology 24:176–187.
Rodriguez, R. N., and Bynum, R. A. (1992). “Examples of Short Run Process Control Methods with the
SHEWHART Procedure in SAS/QC Software.” Unpublished manuscript available from the authors.
Rousseeuw, P. J., and Croux, C. (1993). “Alternatives to the Median Absolute Deviation.” Journal of the
American Statistical Association 88:1273–1283.
Royston, J. P. (1992). “Approximating the Shapiro-Wilk W Test for Nonnormality.” Statistics and Computing
2:117–119.
Silverman, B. W. (1986). Density Estimation for Statistics and Data Analysis. New York: Chapman & Hall.
Slifker, J. F., and Shapiro, S. S. (1980). “The Johnson System: Selection and Parameter Estimation.”
Technometrics 22:239–246.
Terrell, G. R., and Scott, D. W. (1985). “Oversmoothed Nonparametric Density Estimates.” Journal of the
American Statistical Association 80:209–214.
Tukey, J. W. (1977). Exploratory Data Analysis. Reading, MA: Addison-Wesley.
Tukey, J. W., and McLaughlin, D. H. (1963). “Less Vulnerable Confidence and Significance Procedures for
Location Based on a Single Sample: Trimming/Winsorization 1.” Sankhya, N Series A 25:331–352.
Vännmann, K. (1995). “A Unified Approach to Capability Indices.” Statistica Sinica 5:805–820.
Vännmann, K. (1997). “A General Class of Capability Indices in the Case of Asymmetric Tolerances.”
Communications in Statistics—Theory and Methods 26:2049–2072.
Velleman, P. F., and Hoaglin, D. C. (1981). Applications, Basics, and Computing of Exploratory Data
Analysis. Boston: Duxbury Press.
Wadsworth, H. M., Stephens, K. S., and Godfrey, A. B. (1986). Modern Methods for Quality Control and
Improvement. New York: John Wiley & Sons.
Wainer, H. (1974). “The Suspended Rootogram and Other Visual Displays: An Empirical Validation.”
American Statistician 28:143–145.
Wilk, M. B., and Gnanadesikan, R. (1968). “Probability Plotting Methods for the Analysis of Data.”
Biometrika 49:525–545.
Wright, P. A. (1995). “A Process Capability Index Sensitive to Skewness.” Journal of Statistical Computation
and Simulation 52:195–203.
Zhang, N. F., Stenback, G. A., and Wardrop, D. M. (1990). “Interval Estimation of Process Capability Index
Cpk.” Communications in Statistics—Theory and Methods 19:4455–4470.

Proc Capability - Sas User Guide

Uploaded by

Proc Capability - Sas User Guide

Uploaded by

®

Overview: CDFPLOT Statement . . . . . . . . . . . . . . . . . . . . . . . . . . . . 266

Example 6.10: Comparing Goodness-of-Fit Tests . . . . . . . . . . . . . . . 385

Examples: OUTPUT Statement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 449

Graphical Estimation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 535

Introduction: CAPABILITY Procedure

 process capability indices, such as Cp and Cpk

 the ability to analyze data sets with a frequency variable

 the ability to read specification limits from a data set

Table 6.1 Statements for Plots and Specialized Analyses

 Otherwise, traditional graphics are produced by default if SAS/GRAPH is licensed.

Learning about the CAPABILITY Procedure

 “Getting Started: CDFPLOT Statement” on page 267

 “Getting Started: COMPHISTOGRAM Statement” on page 286

 “Getting Started: HISTOGRAM Statement” on page 312

 “Getting Started: INSET Statement” on page 399

 “Getting Started: INTERVALS Statement” on page 427

 “Getting Started: OUTPUT Statement” on page 439

 “Getting Started: PPPLOT Statement” on page 455

 “Getting Started: PROBPLOT Statement” on page 478

 “Getting Started: QQPLOT Statement” on page 509

PROC CAPABILITY and General Statements

Overview: CAPABILITY Procedure

 basic measures of location and variability

 confidence intervals for the mean, standard deviation, and variance

 tests for location

 tests for normality

 trimmed and Winsorized means

 robust estimates of scale

 quantiles and related confidence intervals

 extreme observations and extreme values

 frequency counts for observations

 percents of measurements within and outside specification limits

 confidence intervals for the probabilities of exceeding the specification limits

 standard capability indices and related confidence intervals

 tests of normality in conjunction with capability indices

 specialized capability indices

In addition, you can use options in the PROC CAPABILITY statement to

 specify the input data set to be analyzed

 specify an input data set containing specification limits

 specify a graphics catalog for saving traditional graphics output

 specify rounding units for variable values

 specify the definition used to calculate percentiles

 specify the divisor used to calculate variances and standard deviations

You can use options in the SPEC statement to

 provide lower and upper specification limits and target values

 control the appearance of specification lines on plots

Getting Started: CAPABILITY Procedure

Computing Descriptive Statistics

title 'Process Capability Analysis of Fluid Weight';

Figure 6.1 Descriptive Statistics

The CAPABILITY Procedure

Basic Statistical Measures

Tests for Location: Mu0=0

Tests for Normality

Figure 6.1 continued

Computing Capability Indices

proc capability data=Cans normaltest freq;

Figure 6.2 Capability Indices and Frequency Table

The CAPABILITY Procedure

Process Capability Indices

Syntax: CAPABILITY Procedure

PROC CAPABILITY Statement

Table 6.2 PROC CAPABILITY Statement Options

Plotting and Graphics Options

Data Summary Options

Table 6.2 continued

Robust Estimation Options

Confidence Limit Options

Table 6.2 continued

CHECKINDICES< (TEST = SW | KS | AD | CVM | NONE > < ALPHA=value >)

proc capability data=Process;

Figure 6.3 Warning Message Printed with Capability Indices

The CAPABILITY Procedure

Process Capability Indices

TEST = SW | KS | AD | CVM | NONE

process capability indices, such as Cp and Cpk

the ability to analyze data sets with a frequency variable

the ability to read specification limits from a data set

Otherwise, traditional graphics are produced by default if SAS/GRAPH is licensed.

“Getting Started: CDFPLOT Statement” on page 267

“Getting Started: COMPHISTOGRAM Statement” on page 286

“Getting Started: HISTOGRAM Statement” on page 312

“Getting Started: INSET Statement” on page 399

“Getting Started: INTERVALS Statement” on page 427

“Getting Started: OUTPUT Statement” on page 439

“Getting Started: PPPLOT Statement” on page 455

“Getting Started: PROBPLOT Statement” on page 478

“Getting Started: QQPLOT Statement” on page 509

basic measures of location and variability

confidence intervals for the mean, standard deviation, and variance

tests for location

tests for normality

trimmed and Winsorized means

robust estimates of scale

quantiles and related confidence intervals

extreme observations and extreme values

frequency counts for observations

percents of measurements within and outside specification limits

confidence intervals for the probabilities of exceeding the specification limits

standard capability indices and related confidence intervals

tests of normality in conjunction with capability indices

specialized capability indices

specify the input data set to be analyzed

specify an input data set containing specification limits

specify a graphics catalog for saving traditional graphics output

specify rounding units for variable values

specify the definition used to calculate percentiles

specify the divisor used to calculate variances and standard deviations

provide lower and upper specification limits and target values

control the appearance of specification lines on plots

Specify the NOTSORTED or DESCENDING option in the BY statement in the CAPABILITY

If the value is missing, the observation is excluded from the analysis.

the output data set specified with the OUTTABLE= option

the tests for normality requested with the NORMALTEST option

percentile definitions controlled using the PCTLDEF= option

computing the mode

assumptions and terminology for capability indices

standard capability indices

specialized capability indices

the number of variables is large

batch processing is used

Cramér–von Mises test

X.1/ X.2/ X.n/

x.1/ x.2/ x.n/

x.1/ x.2/ x.n/