BIOS 521 Lab #0: R Tutorial (EHS)

Matt Zawistowski

This tutorial provides an introduction to the R programming software, including basic commands to read in
your data file and begin an initial analysis of the data. The tutorial is designed to help you become more
familiar with basic commands in R. You will start by simply copy-pasting code to the command line. At the
end of the tutorial, you will write your own code by making simple changes to previous examples.
Installation of R/RStudio You first need to install R and RStudio on your personal computer. Go to
https://cran.r-project.org/ and choose the version of R appropriate for your operating system (e.g. Windows,
Mac). Follow the directions to download and install the most recent version of R. Next you will want to
install RStudio, an optional R workspace which provides nice visual features that are particularly helpful
for new users. RStudio can be downloaded at: https://rstudio.com/products/rstudio/download/. Click on
the (free) RStudio Desktop version and then choose the appropriate version for your operating system. This
website provides additional details on installation: https://rstudio-education.github.io/hopr/starting.html.
Once you have installed both programs, launch RStudio and you are ready to begin coding.
Installing tidyverse The first step is to install the tidyverse package which contains nearly all functions
needed to complete the labs this semester. R packages contain code for functions that are not included with
basic R installation. Packages only need to be installed one time and then simply be re-loaded each time
you begin a new R session. To download and install the tidyverse package, copy and paste the following
code into the R console command line. You will need to choose a mirror for download. Just choose anything
in the US.

install.packages("tidyverse")

In the previous step, you downloaded and installed R code to your local computer. Now you need to actually
run that code in the current R session using the library function. You will need to run this command every
time you start a new R session.

library(tidyverse)

Loading Data The next step is to load the data file for the EHS project into the current R session. First
you need to tell R where the file is located. It would be a good idea to create a specific directory on your local
computer where you store the dataset, corresponding code and result files. You can set the current working
directory using the function setwd and passing the path of the directory. Once the working directory is set,
R will now look for files in this directory and print output to this directory. You will need to replace the
text PATH OF FILE with the actual path where you have placed the data file (the path needs to be in double
quotes).

# Set the working directory containing the dataset

setwd("PATH OF FILE")

NOTE: The pound sign (now known as a hashtag) is the comment character in the R language, meaning
the line is ignored by the computer and simply there to help humans read the code. In the code above, you
can see that I added a comment to remind myself what the following line of code does. Adding comments

1
to your code makes it much more interpretable both to others and to yourself. It can be very frustrating to
return to code you wrote in the past and try to remember what exactly it does. Use comments as a favor to
your future self!
The dataset is in csv format (columns separated by commas) so we will use the read_csv function to read
it into the R session. The dataset is read in and stored as an object that you can name. You can call it
DATA, or ehs_data or Fred. It does not matter what you name it, you just need to be consistent throughout
your code. The read_csv function actually stores the data in something called a tibble which has some nice
properties for data manipulation that we will use later. For now, you don’t need to worry about that. I
have included the option na="NA" to tell R that missing data in this file is coded with “NA”. Other common
options for missing data are -9 or simply a blank entry.

# Read in the dataset using read_csv. Store it as an object called DATA Note
# that my data file is stored in a subfolder called 'Data' off of my working
# directory
DATA = read_csv("Data/EHS_521_lab_data.csv", na = "NA")

## Rows: 1993 Columns: 10

## -- Column specification --------------------------------------------------------

## Delimiter: ","
## chr (4): RIAGENDR, RIDRETH3, DBQ700, SMQ681
## dbl (6): SEQN, LBXNFOS, RIDAGEYR, BMXBMI, LBXGH, PAD680

##
## i Use ‘spec()‘ to retrieve the full column specification for this data.
## i Specify the column types or set ‘show_col_types = FALSE‘ to quiet this message.

When the data is read into R using the read_csv function, it will print to the screen the list of variables
that were stored and their respective data type (e.g. chr indicates categorical/character variables, and dbl
indicates variables stored as a number).
If you are unsure about how to use a function or what the potential options are, you can always learn more
about that function and even see example usage by typing a question mark (?) followed by the function
name. For example: ?read_csv.
Working with Data Once the data has been loaded, the first thing you probably want to do is take a peek
to ensure it looks proper. To do that, use the view function. This will open a separate window in RStudio
that allows you to scroll through the dataset.

# Take a peek at your data

view(DATA)

You should notice that that there are many NA values present in the data, these indicate missing measurements
for an observation.

2
Now that you have peeked at the dataset, let’s learn some functions for working with data. First, let’s
see how many observations and variables are contained in the dataset. The dim function will return two
numbers: the first is the number of rows (samples/observations) and the second is the number of columns
(variables).

dim(DATA)

## [1] 1993 10

In this case there are 1993 samples and 10 variables.

Numerical Variable Next we’ll look deeper into the variables in the dataset by performing an exploratory
analysis consisting of graphical plots and computing descriptive statistics. Let’s start with the Age variable
(called RIDAGEYR in the datafile). Since Age is numeric, either a histogram or boxplot are good ways to
visualize the distribution.

# Plot the Age variable using a histogram, use the main and xlab options to
# provide names for the figure and the x-axis, respectively
hist(DATA$RIDAGEYR, main = "Histogram of Age Variable", xlab = "Age")

Histogram of Age Variable

200
150
Frequency

100
50
0

10 20 30 40 50 60 70 80

Age

3
# Plot the Age variable using a Boxplot
boxplot(DATA$RIDAGEYR)
80
70
60
50
40
30
20
10

In the code above, we specified the variable we wanted to examine using the $ notation. When R sees
DATA$RIDAGEYR it knows that you are referring to the RIDAGEYR variable in the DATA data object.
The distribution of Age can be summarized using descriptive statistics such as the mean and standard
deviation, or the Five-Number Summary. The na.rm=TRUE option tells R to ignore any missing data when
computing the quantity of interest. If there are missing data in the variable and this option is not included,
R will return NA rather than the mean of all available numbers.

mean(DATA$RIDAGEYR, na.rm = TRUE)

## [1] 43.36478

sd(DATA$RIDAGEYR, na.rm = TRUE)

## [1] 20.55558

summary(DATA$RIDAGEYR)

## Min. 1st Qu. Median Mean 3rd Qu. Max.

## 12.00 25.00 43.00 43.36 61.00 80.00

4
Categorical Variable Now, let’s look at a plots and descriptive statistics for the categorical variable gender.
Create a table with counts for each level of the categorical variable using the table function:
# create a table of counts
table(DATA$RIAGENDR)

##
## Female Male
## 1029 964

This table now tells us that the gender variable has two levels (Female and Male) and gives the corresponding
counts in the dataset. If you are interested in the proportions (rather than counts), those can be computed
by passing the counts to the prop.table command as follows:
# create a table of proportions
prop.table(table(DATA$RIAGENDR))

##
## Female Male
## 0.5163071 0.4836929

A barchart is used to visualize counts for categorical variable. R provides fine-tune control over graphics, for
example the colors for the bars, by passing options to the function. Here the col= argument lets you specify
the colors of each bar.
# create a barplot
barplot(table(DATA$RIAGENDR), col = c("red", "blue"))
1000
800
600
400
200
0

Female Male

5
Next, lets look at a different categorical variable, diet quality, the variable named DBQ700. Make the count
table and barplot in the same fashion as we did for the gender variable.

table(DATA$DBQ700)

##
## Excellent Fair Good Poor VeryGood
## 143 471 726 137 333

barplot(table(DATA$DBQ700), cex.names = 0.9)

700
500
300
100
0

Excellent Fair Good Poor VeryGood

# The `cex.names` argument in the `barplot` function controls how large to

# print the names of the levels under the bars. I have made them a little
# smaller than the default so they all fit nicely.

Something should look off about the histogram. The order of the bars does not make sense. The diet quality
variable is an ordinal categorical variable since there is a natural ordering to different quality levels. R does
not know this so it simply plotted the levels in alphabetical order in the barplot, making it difficult to see
patterns in the plot. We can use the ordered function to tell R the correct order for the levels of diet quality.

# order the levels of the diet quality variable, be sure that the levels are
# written exactly as they appear in the data (e.g. 'VeryGood' as one word)
DATA$DBQ700 = ordered(DATA$DBQ700, levels = c("Poor", "Fair", "Good", "VeryGood",
"Excellent"))

6
Now, re-run the table and barplot commands. Notice that the levels for diet are in the natural order. The
barplot in particular becomes much easier to interpret.

table(DATA$DBQ700)

##
## Poor Fair Good VeryGood Excellent
## 137 471 726 333 143

barplot(table(DATA$DBQ700), cex.names = 0.9)

700
500
300
100
0

Poor Fair Good VeryGood Excellent

Basic Data Manipulation We will next look at two basic data manipulation functions. First we will
rename a variable. It can be difficult to look at the variable names that come with the dataset, for example
RIDAGEYR, and remember exactly what it is. Let’s just rename it something simpler, like Age. To do this
we will use the rename function. You should feel free to rename variables in your data to something that
is more informative. As a bonus, your updated variable names will appear in any plots and tables, making
them much more readable.

# rename the Age variable

DATA = rename(DATA, Age = RIDAGEYR)

7
Finally, we will create a new variable. We can create a new variable based on existing variables in the dataset
using the mutate function. Imagine that we would like to indicate samples that are over age 50. We will
create a new variable called Age50 that is 1 if the sample is 50 or over and 0 if the sample is under 50. We
will also make use of the ifelse function which takes a logical condition (e.g.. Age>=50) and returns 1 if
that condition is true for a sample and 0 otherwise. (Remember you can use ?ifelse for additional details
on the usage.)

# create a new variable Age50 to indicate which samples are aged 50 and above
DATA = mutate(DATA, Age50 = ifelse(Age >= 50, 1, 0))

We can see count the number of samples aged 50 and above by creating a table:

# count the number of sample 50 and above

table(DATA$Age50)

##
## 0 1
## 1195 798

8
Lab #0 Questions: Now, it’s your turn to try writing R code on your own. Answer the following questions
by writing code based on the examples from above. Submit both the code you wrote and the answer for
each question.

1. Each dataset contains a variable for self-reported race. Rename that variable to simply race. Is there
any need to order the levels of the race variable? Why?
2. How many levels are included in the race variable? What are the proportions for each group in your
dataset?
3. Compute the mean, standard deviation and Five-Number Summary for the BMI variable in your
dataset.

4. Based on the descriptive statistics for BMI, do you have any concerns about potential outlier values?
5. Based on the descriptive statistics for BMI, do you think the distribution for BMI is most likely to be
left skewed, right skewed or symmetric? Why?
6. Confirm your answer to the above question by creating an appropriate plot to visualize the distribution
of BMI.
7. Create a new variable called LogAge containing the natural logarithm of the Age variable. (HINT:
The log function computes the natural log, logAge=log(Age) )
8. What does the distribution of your new variable logAge? (HINT: Create a histogram or boxplot of
the variable you created in the previous step.) Compare this to the shape of the original Age variable?
That is, how did applying the natural log function change the shape of the distribution of ages in the
dataset.
9. Suppose that you are interested in designing a study with individuals aged 65 and above. How many
such samples in your dataset? (HINT: create a new variable to identify samples 65+ and use a table
to count.)