MANONMANIAM SUNDARANAR UNIVERSITY

DIRECTORATE OF DISTANCE & CONTINUING EDUCATION

TIRUNELVELI 627012, TAMIL NADU

M.Sc. PSYCHOLOGY - I YEAR

DKPY4/DSY14 - RESEARCH METHODOLOGY

(From the academic year 2016-17)

Most Student friendly University - Strive to Study and Learn to Excel

For more information visit: http://www.msuniv.ac.in

 SYLLABI.BOOK MAPPING TABLE
Research Methodology

Syllabi Mapping in Book

UnitI
Meaning of Research: Objectives, t)?es, approaches, significance of Unlt 1: Meaning of Research
research, methods Vs methodology, research process, flow chart, criteria (Pages 3-24)
of good research, problems encountered by researches in India.
Measurement Scales: Nominal scales, ordinal scales, interval scale,
ratio scale, source of errors in measurement, tests of sound measurement,
test of validiry test of reliability, test of practicability.

UnitII
Meaning and Definition of a Research Problems: Selection, sampling, Unit 2: Meaning and Definition of a
steps, t5pes, sample size, testing hypothesis-I, (parametric or standard Research Problem (Pages 2568)
test of hypothesis), testing hypothesis-II (non parametric or
distribution), free tests.

UnitIII
Sempling Design: Census and sample survey, implications of sample Unlt 3: Sampling Design
design, steps in sampling survey, types of universe, sampling unit, (Pagec 69-lt4)
source list, size of sample, parameters interest, budgetary constraint,
sampling procedure, criteria of selecting a sampling procedure,
characteristics of a good sample design, different tlpes of sample design,
how to select a random sample, random sample from an infinitive universe
complex random sample design.
Sampling Fundamentals: Need for sampling, some fundamental
definitions, important sampling distributions, central limit theorem,
sampling theory Sandier's A test, concept of standard errot estimation,
estimating the population, mean, estimating population proportion,
sample size through the approach based on precision rate and
confidence level, deterministic sample size through the approach based
on Bayesian statistics.

UnitIV
Research Design: Meaning ofresearch design, need for research design, Unlt 4: Research Design
features of a good design, important concepts relating to reseaich (Pages 115.165)
design, different research designs, basic principles of experimental
design, conclusions, developing research plan.
Experimental Designs: Between group designs, within group designs,
mixed designs, latin square designs.
Non experimental Design and Correlational Methods: Non and quasi-
experimental designs, correlational designs, newer social methods,
advanced correlational methods, discriminant function analysis.
Qualitative Methods: Definition and aim of qualitative research,
construction of reality, subjectivity, dynamic research, process of
documentation of qualitative research.

UnitV
Computer and Its Role in Research: Computer applications, the Unit 5: Role of Computers in Research,
computer system, important characteristics, the binary number system, lnterpretation and Report Writing
computers and researchers. (Pages 167-189)
Interpretation and Report Writing: Meaning, reasons, techniques,
precautions, steps in report writing, layout, types, oral presentation
and precautions in report writing.
 CONTENTS
l\'t'R0DI-r('Tr0N
I
UNIT I IVTEANING OF RESEARCH
3-24
l.() lntroduction
l.l Unit Ob.icctivcs
l.l lVlcanins ol' Rcscarch
1.3 { )h.icctivex ol Rcscarch
1.4 Typcs of Rcscarch
1.,5 Approache.s to Rescarclr
1.6 Si_snificancc of Rcsearch
1.7 Method vs. Mcthodology
L8
Rcscarch Prclcess
1.9 Flow Chart
l.l0 Criteria of Good Rcscarch
I.lI Problcnx Encountcretl by Inclian Rcscarchcrs
I.l2 Mcasurcrrrcnt Scales
I.l3 Sourccs ol' Error in lVlcasurcmcnt
l.l4 Test ol'sourrd lUcasurcnrcnt
l.l.l. I 'l cst ol-\ralitlit-v..
l.l.l.l Tcst ol'Rcliability
I'14.-1 Tcst ol'l)racticalilv
l.l5 Sunrrnary
l.16 Key Tcnns
l.l7 Answers to 'C'hcck Your progrcss'
l.l tl Qucstions and Exerciscs
l.l9 Furthcr Reading

UNIT 2 MEANING AND DEFINITION OF A RESEARCH PROBL},M

25.68
2.0 Introduction
2.1 Unit Objcctivcs
2.2 Research Problem and Working Hypothesis
2.3 Sarnple Selection. Types and Size
2.4 Paranretric and Non-parametric Tcsts
2.5 Testin_r f lypothcsis I: paranretric or Standard'fc.stin-r
2.6 Testing Ilypothcsis Il: Non-paranrctric -fcsting
2.7 Corrclalitn and Rcgrcssion
2.['l Surnrnary
2.9 Kcy Tclrr:s
2.10 Answers to 'Check your progress,
2.ll Questions and Exercises
2.12 Further Reading
UNIT 3 SAMPLING DESIGN 69-114

3.0 [ntroduction
3.1 Unit Objectives
3.2 SamplingDesign
3.2.1 Steps in SamPling Designing
3.2.2 Principles for Selecting a Sampling Procedure
3.2.3 Systemic Bias and Sampling Errors
3.2.4 Criteria ofGood Sample Design
3.2.5 Types of SamPling Designs
3.2.6 Types of SamPling
3.3 Statistical Distributions
3.4 Surmnary
3.5 Key Terms
3.6 Answers to 'Check Your Progress'
3.7 Questions and Exercises
3.8 Further Reading

uNrT 4 RESEARCH DESIGN 11S165

4.0 Introduction
4.1 Unit Objectives
4.2 Research Design
4.3 Experimental Design
4.3.1 ImportantExperimantalDesigns
4.3.2 Quasi-ExPaimentalDesigns
4.4 Qualitative Methods
4.4.1 Definitions of Qualitative Research
4.4.2 Distinctions betrveen Qualitative and Quantitative Research
4.4.3 ResearchPersPective
4.4.4 Texts and Realities in Qualitative Research
4.4.5 Theories in Qualitative Research
4.4.6 Sampling Strategies in Qualitative Research
4.4.7 Types of Interviews for Qualitative Research
4.4.8 Focus GrouPs
4.4.9 VisualData
44.10 EthnograPhY
4.4.11 CodingandCategorizing
4.4.12 ContentAnalysis
4.4.13 Documentation ofData in Qualitative Research
4414 Presenting the Findings/Results Following the Qualitative Research
4.4.15 Computers in Qualitative Research
4.5 Summary
4.6 Key Terms
4.7 Answers to 'Check Your Progress'
4.8 Questions and Exercises
4.9 Further Reading
UNIT 5 ROLE OF COMPUTERS IN RESEARCH, INTERPRETATION 167.189
AND REPORT WRITING
5.0 Introduction
5.1 Unit Objectives
5.2 ComputerApplications
5.3 The Computer System
5.3.1 ImportantCharacteristics
5.4 The Binary Number System
5.4.1 Decimal toBinaryConversion
5.4.2 Binaryto Decimal Conversion
5.4.3 Computations in Binary System
5.5 Computers and Researchers
5.6 Interpretation and Report Writing
5.6.1 Meaningoflnterpretation
5.6.2 The Need and Importance oflnterpreting the Findings
5.6.3 Techniques oflnterpretation
5.6.4 Precautions during Interpretation
5.6.5 Steps in ReportWriting
5.6.6 The Layout ofthe Report
5.6.7 Tlpes of Reports
5.6.8 Oral Presentation ofReports
5.7 Summary
5.8 Key Terms
5.9 Answers to 'Check Your Progress'
5.10 Questions and Exercises
5.ll Further Reading
 lnl n ulut.'t itttt
INTRODUCI-ION
Dttingfc.tcttr(:ll(lllllr: l'litllt'rrrsill(lupi(lclltirrIrU] rrl'psyclyrlictlisprtle.rstbnlsan
ilttptrrl;trtt Pilrt ill tltr: sluth'ol psvcholt)-{\. lltc lrook /r'clrc,crn,h llctlutcfutlr4it.
NOTES
disr"-usscs tlilli.'r.'nt rcs,.:itrch techniqucs starting linrn sanplin-q and rcscarch
dcsign
to ditlbrenl slirt rst ical lcclrrriqucs.
I Irrit I tlrscrrsst's lhc rncarrirrg. dctjnitiorr lurd \c()l)e olresearch. Thc unit
also cxplairts t ltc lrr rics o l'rcscarclt l)r()cc\s. tlrllL.r'c1t tr pg5 11t'1a.car cI alcl thc
di lll'rcncc hcl r, cctt rcscllrclt rlrcl l rptls u rrtl rlcl lrptlg logr.:

[Jrlit ] lltt'on s ltgltl tttl ltou to rlcl'inc a rcscLu'clr ltnrhlcrn antl lirnrrrrlatc a
hyxrthcsis. l-ltt unit ltrflhcrcxplairslriu.anrctr.ic arrln()n-p:l.lultclrrc test:llrlgur.lcs
lx^r'to ch.osc ditll'rcrrt kinds .t'tcsts tti tcst a hrp.thcsis.
Unit 3 csplainstlu conccpt ofsanplin-g dcsi-un aril cril]e\rc,t twcs ol'vunpling
proccdurcs. Tlrc turit also cxprlains probabilitv ancl rron-llrohabilitv .u,r.,1ilirg
tcclmitlttcs. dil'llrcnt trpcs ol'sarrtpling cmors ancl horv to nrirriprizr-- thcnr.
l..rnit'l cxpltlrcs tlrc ctutccpl ol'rcsearch clcsign and tliflbrr.ut t)'pcs ofrcseadr.
It also discttsscs lrorv to dilll'r'crrtratr'lrctu'ccn the qualitatir,e and cluantitativc
rcscarch.
Llnit 5. erittttittcs httrr cotttprrlcrs lrar c rcr,olrrtitlnizctl tlrc rcscru.ch rr,<lrk.
Thc trnit alstt crplailts horr lo rvritc ir rcscurclr r-!.p()11. irrtcrpret thc rcsults alcl
drarv corrclusr()ns.
Each unit. in this lrook. is supplcnrcrrtccluith Sururrur KcyTcnns. Arurrcru
kt 'Chcck lirul p11-'1'gss'.
Qucslions attd [:rcrcis.'s and [turthcr Rcaclin-t scctions.
to aitl tlrc rcardcr.

Sr ll-l nrttltt. tu rtu I .l kt t.1, td I

 Meaning of Research

UNIT 1 MEANING OF RESEARCH

Strucfure
NOTES
1.0 Introduction
l.l Unit Objectives
1.2 Meaning of Research
1.3 Objectives of Research
1.4 Types of Research
1.5 Approaches to Research
1.6 Significance of Research
1.7 Method vs. Methodology
1.8 Research Process
1.9 Flow Chart
l.l0 Criteria of Good Research
l.ll Problems Encountered by Indian Researchers
L12 Measurement Scales
l.l3 Sources of Error in Measurement
l.l4 Test of Sound Measurement
l.14.l Test ofValidity
1.14.2 Test of Reliability
1.14.3 Test of Practicality
1.15 Summary
1.16 Key Terms
Ll7 Answers to 'Check Your Progress'
l.l8 Questions and Exercises
l.l9 Further Reading

I.O INTRODUCTION

In layman's terms research means the search for knowledge. Scientific research is
a systematic and objective way ofseeking answers to certain questions that require
inquiry and insight or that have been raised on a particular topic. The purpose of
research, therefore, is to discover and develop an organized body ofknowledge
in any discipline.

Research is a journey of discovery. It is a solution-oriented inquiry that

must be objective and repeatable. It should inspire and guide further studies and
should foster applications. Research will provide practical benefits ifit can provide
advancedunderstanding ofa discipline or suggest ways to handle some situations
thatweconfront.
Scientific research involves contro lled observations, analysis of empirical
data and interpretation offindings. This can further lead to the development of
concepts, generalizations, etc., on the basis ofwhich theories could be formulated.
Such an investigation could help in determining cause and effect relationship. The

Sell:lntr*"rtoro, Material 3
,lkottirrg rtl Rr', r,r', i: ttllitrt.tlr' aittt of'\()ci.ll sc,cnec rcsc.lcll is thc control and prccliction of
[thar,'iour.
Knowinq hon fo do research: Every rcsearclrcr should have thc nccessary
traitrirr;: in i:.irlhcrinq,i.rtrr. orsilltizurg ntalcrials suilablr anrlciruasirrg in l'icklor
NO'T[,S
latxtttlttry w ork. ;t: I'crlur ctl. I lc slroukl also havo tlu coupctcncc in usinu statistics
lirr trcating thc'datu and the abitity to intcrpret thc'data collcctcrl nrcaninufully.
'l raining of the mind:
Rcsc,arch nccds disciplinc. right nrcntalnukcup. tl'rc
abilitl'to manage tinrc cflbctivch: oh.icctir itr,. logicalthirrhing, thc capac:ity to
ovaltntc tltc rcsults o{'tlrc tcscarc,h and abilitv to carctirllv arist..\\ rhe findings tlrat
arc founcl bv thc rcscarch.
Rcsc:trch tlittn allon s pcoplc 1o tnakc inliruncd dccisiorrs by cxtrapolating
thc fitrdnrgs lirrttt thc ticklor latxrraton'on to rcallill'situatiorrs. I lris is thc practical
application ol'thc tindings scncriltc(l b1, rcrcalch.
Rcscarch is also a way ofprcparing thc nrinrl to look at things iu a liesh or
dillbrcnt way. out of such on oricntation n'ould comc ncw and innovativc
obscnations ahout cvcryday cvcnts anclhappcnings. This is how originalitycomcs
about itt rcscarch. Sotrtc ofthc nrost outstanding discovcries have bccn made in
thc rnost scrcndipitous manner. Somc outstandin.q results havc been obtained by
rcseatchcrs who had kepl their minds open and fi'cc ofcluttcr. This otabled thern
to scc slafl lingly' rrc'* conncctiorrs.

l.l UNIT OBJECTIVES

Aftcr -rroing tlrrough this unit, you r,r'illbe ahlc to:
o Explain the rncaning ofrescarch
o Lcarn thc ob"icctives and types ol'rescurch orientations that arc possible
o Examine the significance olrcscarch
. Kltttu thc distinctir.ln hetlr,cstt rcscurclr nrcllrotls and ntcthodology
o (iain insights into thc rcscarch proccsscs

1.2 NIEANING OF RESEARCH

Rcsciu'clt itt cutrutton parlatrcc rcfbrs to scarch fbr knowlcd-uc. Orrc can also dr:fine
rescarch as a scientific attd svstcttratic serarch for pcrrlinurt inlbnrratiotr on a spccilic
topic. In lirct. rcscarch is att at1 ofscicntific irl'cstigation. Accorclirry to the Advanccd
Leamcr's Dictionary ofCurrettt English, 'rcsearch is a carcful invcstigation or
enquiry cspeciallv a thorough search for ncw tacts in any branch of knowledge.'
Redrnan and NItlry ( I 923) deftred research as a 'svstc-rlatizcd eflort to gain nerv
knowlcdgc.' Sornc pcople ctxrsitlcr rescarch as a \oyage ofdiscoverythat involves
rnovcnrcnt lionr thc known to the unknow'n.

4 Sel/:ln.\n1!di(uiul lfuttrtd
 Research in a technical sense is an academic actirity. Clifforcl Wtrocly defined Meuning of Research
research as an activit,v tlmt cottrpriscs dctirring ancl redelining problems. tirmrulating
a hypothesis; collccting. organizing ancl cvaluatirrg dltta; ntaking clctlucticlns and
reaclring conclusions: and citretirllv tcrsting thc conclusit>ns to rte'tcnnine ifthey
sr.lpport the formulatecl lrypothesis. D. Slcsinger and N,t. Stcphcnson, in the NOTES
Eno'ck4xrcdia ol'&x'iul Sc'ie'ne't'.s. dclinctl rcsearch as 'the nunipulatisn ofthings,
concepts or synrhols lirr thc purp()sc ol'gencralizing. extcncling. correctinq or
verifring thc'knowlcdgc. wlrcthet that knowlcrlge aids ir tlrc corstruction oftheory
or in the practice ofan afl.' Ilcsearch is thus an original contribution to thc existing
stock of knowledge rnaking for its advancement.
Principles ofa tyrrical rcsearch:
o It is basccl an utrpriricltl clalir
o [t i,vol'cs prccisr,, tltrse^,ati..s irntl rrrc:.rsrr.cl]..,t\
o [t is aimed at developing thcories, principlcs and gcnc,mlizatigns
o There arc systcmatic. logical trrroceclurcs inr,oh cd
o It is rcplicable-
o The findings ofthc rcscarch uced to be reporlctl
A flow chart in tenns ol'how rcsearch prccecds is prcscntcri in Section 1.9
for better understandin-s rvhere thr: c:ritc-ria ofgoocl rc.searclr hal.c becn i6entified
in easy steps.
Finally, one is exposed to some ol'the specilic problons tllrl rcscarclgs in
India have to face. Measurenrent scalr,'s lrc irrtxrducctl ti,r r:it l.:i:nclit ol'the
researcher in turtns ol'the ir lttiltlrsnla tica I propcrtics. 'l'lrc nt.r
l(,r :cirhs t 1ar ftlve
been widely uscd havc bccrr 1lr'.'scnlcrl irnrl rrisr:rrssci.l ilr s,r,r.: .l,.i,rr! r'lr ..:.rlcs
arc:
o Nominalscalc
o Ordinalscalc
o lntcrval scalc and
o Ratio scale
Their differences and significance in teffns of use, sources of eror in
measurement have been dealt with in detail the following sections so as to facilitate
precise measurements in research rn'ork.

The.qualities of good or sound measurement have been describecl for the

researcher to be aware of. The three major tests of effective measurements
are
specified and discussed at length. These arc tests of reliability. r,alidity, and
practicality. I

S,ll:lutrru<tionul llute,rial 5
Meaning of Research
1.3 OBJECTIVES OF RESEARCH

The objective of any research is to find answers to questions through the

NOTES application of scientific procedures. The main aim of any research is exploring
the hidden or undiscovered truth. Even though each research study has a specific
objective, the resemch objectives in general can be categorized into the following
broad categories:
o Exploratory or formulative research studies: These are aimed at gaining
familiarfu with a particular phenomenon or at gaining new insights into it.
o Descriptive research studies: These me aimed at accurately portraying the
characteristics of a particular event, phenomenon, individual or situation.
o Diagnostic research studies: These studies try to determine the frequency
with which something occurs.
o Hypothesis testing research studies: These studies test a hypothesis and
determine a causal relationship between the variables.

1,4 TYPES OF RESEARCH

The basic tlpes ofresearch are as follows:

o Descriptive vs. analytical research: Descriptive research includes
different kinds offact-finding inquiries and surveys. The main objective of
this research is describing the state ofconditions as it exits at the present
moment. For descriptive research studies in the social sciences, we often
use the term ex post facto research method whose main characteristic is the
researcher has no control over the variables; he can report only what has
happened or what is happening. Used often for descriptive studies, most
ex-post-facto research projects, seek to measure such phenomena as
preferences ofpeople, frequency of shopping, or similar data. The methods
used in descriptive research include all kinds ofsurvey, comparative and
correlation methods. On the other hand, in analytical researctL the researcher
uses the information and facts alreadyavailable, and anallzes themto make
a critical evaluation.
o Applied vs. fundamental research: Research can be either fundamental
(basic or pure) or action-oriented (applied) research. Fundamental research
focuses on finding generalizations and formulating theories whereas applied
research aims at finding a solution for an immediate problem facing a society
or abusiness/industrial organization Gathering knowledge for knowledge's
sake is termed pure or basic research (Pauline V. Young). Research studies
carried out witha view to making generalizations about humanbehaviour
are examples of fundamental research. Research aimed at particular

6 Self-lnstrrctional Mateiql
 concfusion (for a particular problem). Doing research on a current social or Meaning of Research

business problem is an example ofapplied research.

o Quantitative vs qualitative: Quantitative research is based on the
measurement ofquantity or amount. This research can be applied to the
NOTES
phenomena that can be quantified. on the other hand, qualitative research
focuses on qualitative phenornena such as happiness, beauty, etc. that cannot
be strictly quantified. 'Motivation Research' related to hurnan behaviour is
a ffi ofqualitative research.
o Conceptual vs empirical research: Conceptual research is concerned
with some abstract theory or idea(s). Philosophers and thinkers generally
use the conceptual research for developing new concepts and for
reinterpreting the existing. Enpirical research, on the ottrer hand, relies only
on real experiences and observations. It is data-based research and its
conclusions canbe verified byobservations or experiments. It is also called
experimental type ofresearch. In empirical research all facts are obtained
at first hand, at their source, and at times by stimulating the production of
desired information. To prove a given hypothesis, the evidence gathered
through empirical studies and experiments is considered to be the most
powerful and accurate.
r some othertypes of rcsearch: All other different types ofresearch are
variants or combinations ofthe approaches discussed above. Research can
also be classified based on the conditions in which research is carried out,
for exanple, the purpose ofresearctq the environment inwhichresearch is
done, the time required to accomplish research, etc. From the time point of
view, research can be regarded as either as one-tirne or longitudinal research.
If the research is confined to a single time-period, it is called one-time
research whereas the research carried over several time-periods is
considered as longitudinal research. Depending upon the environment in
which research is to be carried out, it can be termed as laboratoryresearch
or fieH-setting rcsearch or simulation research Clinir:al or diagnosic research
involves using case-study methods or in-depth approaches into the causes
or events. clinicalor diagnostic researchmethods also use verydeep probing
data gathering devices and very small samples to obtain the necessary data.
Historical Research method uses historical sources like documents,
archeological remains, archives, old informatiorl etc., to studypast events
and ideas, including the ideas, thoughts, philosophy ofpersons and groups
at distant time periods. Conchsion-oriented and decision-oriented research
are two other classifications, where in conclusion-oriented research, a
researcher is free to pick up a problerq redesign the enquiry as he proceeds
ard conceptualize as he sees fit. In decisbn-oriented researctu the researctrer
to take the decision-rnaker into confidence and take his suggestions at every
stage ofhis research. It has to be much focused.

SeA-hslruclional Moterial 7
Meaning of Research
I.5 APPROACHES TO RESEARCH

Quantitativc ilppn)llch and clualitutivc approach are thc tu'o basic approachcs to
NOTES rcscarch. Thcsc ttr o ;xtt'adirtn.r r.lrc basctl on t\\,o dilllrcnt and cornpcting rvays
o1'undcrstanding tltc world. Tltr.rsc conrpcting rvuys ol'corrrprchcnding thc rvorld
arc rcllsctecl in thc \\'a)'thc rcscarch elatir i.s collcctccl (c.g.. rvords vs nur)lbcrs),
and the pcrspcctivc ol'thc' rcscarclrcr lpcrspectival vs objccrivc). 'fhc perspcctilcs
of thc' participants arc vcrv critical.
(i) Quantitative approach: Il'there has hccn ont. ovenl,hclming conscnsus
among acldctnic prs"tholouists on a singlc poirr or cr thc pi.tsl ll's clecaclcs,
it is tlrlrt thc b!.st err rpirrclrl r cscu clt in tlrc li-.ltl is lirlrlr r-lrotrrtlcrl fur t1r rllrl itlt irc
tlletlttltls. lltlirtrltlrllnraclt.clatlrisuiinr'lllr(l intlrr.,r,irtrrli.,t'lirrrn urrrl tlrut
tltltldatuissttlrlcclctl 1(,rigorous(luiultit:ilt\c-utill\:!rt!t,trrgrrl lrrrtl 1rt'1Hl
litsltittn. lnlL'rcntial. c.xpcnntcntalarrd .irrrrrlatiorr:rirplolrclres tu'c thr] sub-
classifications ol'quantitltil'r.. approriclr. lrrli'r.cntlalrPpro:.rch ltr rcscarch
ltlcuses on sttr\c.v rcseurclt u'ltcrc r.latltrlscs ar,-. brrilt stud!ing sanrplcs rll'
populatiorr and thc'n thcsc tlatabascs are usctl to inll'r characteristics or
relationships in populations. IIt experimental approach. greater control is
exercisccl o\cr rhc research cnr,'ironmcnt and oftr'n. sorne variablcs
( independent ) I ariatrlcs are controlled or manipulated to record their c'ffi:cts

on othorvariables (depcnclr-'nt variables). In simulation approach. artificial

enviroruncltt is r:rrnstructcd u ithin which relevant data and information is
gencrated. 'rhir rva-l'. the dyrrarnic behaviours clf a system arc observed
undsr c()11ti (jlI I contlititlns.
(ii) Qualit:ltirt'rl1)ttrolch: I'!tis lpptrurch lo rcscarch is concctrrcd n'ith
rtrh.ier.tirc t\i\'\:ltlctttrtluttittrtle\. rrltinionianclhclraf iour. [tcscarclt in
sttclt it rtttrtttiotr ts I lintclrorr ul'rcscl.rt'chcr":, irrrrsltt irtttl irttprcssiotts. Sttch
ittr irlrPrt i:tcll l\ ) t'c\c:lt'(ll t:L:l)ct i"ll,j\ l'r.\ult\ cith.'r irt I tr )n-.lui.ullititl i\.!. li)fllf Or
itt tltC lirntts tt llre lt iu !'lt(rt srrlrjectcrl l() t'i.U(ttr,us (lrtiulltlrtlt\c ltrt.tlf,ris.
Ilelow' table gir cs t)'pcs of rcsearch. rnethods crnployed und tcchniquos
used by those types ofrcscarch.

CHrcxYoun PnocRnss
l. What is scientific research'J
2. Dillerentiate betwecn fundamental and applied research.
3. Differentiate between conceptual and empirical research.
4. Differentiate between quantitative and qualitative approach.

8 Self-lnsntutionul l,laterial
 Table 1.1 Types ofResearch Meaning of Research

Type Methods Techniques

I Library (i) Analysis of historical Recording of notes, content analysis,
records tape and film listening and
(ii) Analysis of documents manipulations, reference and abstract NOTES
guides, content analysis.
2. I Fintd (i) Non-participantdirect Observational behavioural scales, use
I R"rrorrh obsenation ofscore cards, etc.
(ii1 Participantobscrvation Interactional recording, possible use of
tape recorders, photographic
techniques.
(iii) Mass observation Recording mass behaviour, interview
using independent observers in public
places.
(iv) Mailquestionnaire Identification of social and economic
background of respondents.
(v) Opinionnaire Use ofanitude scales, projective
techniques, use of goniometric scales.
(vi) Personal interview lnterviewer uses a detailed schedule
with open and closed questions.
(vii) Focusedinterview Interviewer focuses attention upon a
given experience and its effects.
(viii) Group intervierv Small groups ofrespondents are
interviewed simultaneously.
(ix) Telephone survey Used as a survey technique for
information and for disceming opinion;
may also be used as follonup
questionnaire.
(x) Case study and life Cross-sectional collection of data for
history intensive analysis, longitudinal
collection of data of intensive

,i Laboratory
Research
Small group study of random
behaviour, play and role
character.
Use of audio-visual recording devices,
use ofobservers. etc.
analysis

1.6 SIGNIFICANCE OF RESEARCH

Research involves developing a scientific temper and logical thinking. The
significance ofresearch-based answers can never be underestimated. The role of
researchis speciallyimportant inthe fields ofEconomics, Business, Govemance,
etc. Here research helps in finding solutions to problems encountered in real life.
Decision-making is facilitated by applied research. Research is also of special
significance in the operational and planning processes ofbusiness and industry.
Here logical and :nalytical techniques are applied to business problems to maximize
profits and minirnize costs. Nlotivational research is another key tool in
understanding consumerbehaviour and health related issues. Reqponsible citizenship
concerns can allbe addressed through good research findings. Social relationships
involving issues like attitudes, interpersonal helping behaviour, environmental
concefiN like crowding, crime, fatigue, productivityand otherpracticalissues are
all capable ofbeing addressed well by scientific research.
SeV-lnstructionalMaterial 9
Meaning of Research Social science research is extremely significant in terms ofproviding practicai
guidance in solving human problems ofimmediate nature.
Research is also important as a career for those in the field of academics. It
could be a career option for profbssionals who wish to undefi ake research to gain
NOTES
new insights and idea generation. Research also fosters creative thinking, and nerv
theorizations.
Research for its own sake and forthe sake ofknowledge and for solving
different problem is all require fonnal training in scientific methodology.

1.7 METHOD VS. METHODOLOGY

Research methods: They refer to all the methods the researchers use while
studying the research problems and while conducting research operations. In
general, the research methods can be categorized into the following three groups:
(i) The first group includes the methods that are concerned with the data
collection.
(u) The second includes the statistical techniques needed for mapping
relationships between the unknowns and the data;
(iii) The third group contains the methods necessaryto evaluate the accuracy of
the results obtained.

Research methodology: It is the procedure that helps to systernatically

proceed in steps to solve a research problem. Research methodology is a broader
concept that includes not the research methods;but also the logic behind the
research methods in the context of a particular research study; and it explains the
reasons for using particular research methods and statistical techniques. Research
mythology also defines how the data should be evaluated to get the appropriate
results.

CnscxYouR PRoGRESS

5. What is the significance ofresearch?

6. Diflerentiate between methods and rnethodology.
7 . What are the criteria for a good researchs?

1.8 RESEARCH PROCESS

Research process includes steps or a series of actions and logical sequence of

those steps to carry out research effectively. The various steps in a research process
are not mutually separate, exclusive or discrete, but they at the same time need not
t0 Self-lnstructional Material
ahvays follow each other. The researcher, at each step, anticipates subsequent Meaning of Research
steps and the requirements. The tentative order ofthe steps and the procedural
guidelines ofthe research process are as given below:
(i) Formulating the research problem: At the very beginning ofresearch,
NOTES
the researcher must clearly define the research problem, i.e., the area of
interest, the matter to be to be inquired into, etc. The problenl before being
solved, is initially stated in a broader perspective and then the researcher
arri',es at the specific questionbygraduallyreducing the ambiguities, ifany.
Then, immediately after fonnulating the problenl the feasibility ofdifferent
solutions is studied before choosing the right solution.
(ii) Extensive literature survey: After formulating the research problern a
brief summary of it should be prepared-this is an essential step. while
writing a Ph.D. thesis the researcher has to prepare a synopsis ofthe topic
and submit it to the appropriate conrmittee or research board for approval.
Synopsis preparation needs extensive survey ofthe literature connected
withtheproblem.
(iii) Development of a working hypothesis: After surveying the literature,
the researcher should clearly state the working hypothesis, which is a tentative
assurnption made before testing it in logical or empirical sequences.
Hypothesis must be as specific as possible and should be limited to the
intended research. This helps to choose the right process.
(iv) Preparing the research design: The next step, after clearlydefining the
research problerrl is preparing the suitable research design. The research
design inchrdes the conceptual framework within which research would be
carried out. A good and planned research design helps to carryr out the
study in an efficient rnanner saving time and resources. It helps to gather the
most useful infonrlation and assists in arriving at the accurate results. Sinrply
put, a good research design facilitates the collection ofrelevant evidence
with minimal expendirure ofmoney, effort, time and other resources.
(v) Determining sample design: Auniverse or population includes allthe
items under inqury. Ifall the items in the population are inquired then such
an inquiry is called census inquny. In a census survey, all the items are
covered and so the highest accuracy is obtained. But this may not be
practicable in surveys involving a big population. census surveys need huge
amounts oftime, money and energyl Hence, quite often it iswise to select
only a fe'*' items from the universe for study purposes. Technically, such a
smati ;;nd convenient number ofitems selected, is called a sample. Specified
plan olthe size and method ofcollecting the sample is technically known as
sample design.
(vi) Collecting the data: In most cases" the data at hand is insufficient and
there is always of fresh data. There are difftrent ways ofcollecting
a need
the appropriate data which differ considerably in terms of relevance,
Self-Instructionol Material 1l
Meaning of Research expenditure, time and other resources. Therefore, the researcher must select
the most appropriate method ofcollecting the data after considering objective
ofthe research, the nature of investigation, time and financial resources
available, scope ofthe inqury, and the desired degree ofaccuracy.
NOTES
(vii) Execution ofthe project:This is an important step in the research process
because ifthe executionproceeds on the correct lines, the collected data
would be dependable, adequate and accurate. Therefore, systematic and
timely execution ofa project plays a crucial role in ensuring the right results
at the end.

(viii) Analysis of data: After collecting the data, the next step is analyzing the
data. The data analysis includes a number ofcloselyrelated operations like
speciSing different categoriesofdata, differentiating andtabulating the data
into different categories, applying the statistical techniques and formulae to
the data, doing the right calculations and then drawing statistical inferences.
Various tests, such as chi-square test, f-test, F-test, etc. he$ in data analysis.
(ix) Hypothesis-testing : Aft er analyzngthe data, the researcher should test
the working hypothesis against the statistical inferences obtained after
analyzing the data. The question now should be answered: Do the findings
support the working hypothesis or they contradict?
(x) Generalizations and Interpretation: If a hlpothesis is tested and upheld
sufficient number of times, the researcher can arrive at a generalization.
The degree of success of a research is calculated on the basis of how
much the arrived generalizations are close to the acceptability. If the
researcher starts with no hypothesis, the researcher will interpret his
findings on the basis of some existing theory and this is known as
interpretation. The process of interpretation often triggers new questions
which lead to further researches.
(xi) Preparation of the report or the thesis: Finally, the researcher has to
prepare the report ofwhat has been studied. Report must be written with
great care keeping the following lay'out inmind:
a. The preliminarypages: The preliminarypages ofthe report should
contain the title, the date, acknowledgments, foreword, table of
contents, list oftables, list ofgraphs and charts (ifany).
b. The main text: The main text ofthe report should have introduction,
sumnrary offrrdings, main report, conchrsion and suggestions for future
research.
c. The closure:At the end ofthe report, appendices shouldbe listed in
respect o f all technical data, fo llowed by biblio graphy. Index terms
should also be given specially in a published research report. A11
references should be cited as per the research writing formats.

12 Self-lnstructional Materiol
 Meaning of Research
1.9 FLOW CHART
RESEARCH PROCESS IN FLO1V C}IART

NOTES

Where @ = iecd back (I{elps in controlling the sub-system

tlr which it is transmittcd)
@=recdr'onvard(;:jrj;i".:H,:.ff
Ji1l,i,l,."r

Fig. 1.1 Research Process

In figure 1.1 the flow chart indicates the sequential steps to be followed in
the research process. One must start with defining the research problem along
with reviewing the relevant literature in the field to become frmiliar wittl the corrcepts
and theories relevant to the issue to be investigated. The next step is the formulation
ofthe hypothesis, which is followed by the research design and sample selection.
Then the collection of data and its analysis is to be attempted. After that the
interpretation and the report writing stages complete the research report. These
have to be written step by step and then edited and refined several times before
preparing the final report.

1.10 CRITERIA OF GOOD RESEARCH

Whatever be the type of research one undertakes, certain common criteria of
good scientific methods have to be followed. Agood research follows logical
methods, is systematic, and structured in accordance with well defined sets of
rules and practices to enable the researcher in arriving at dependable conclusions.
Botlq deductive reasoning and inductive reasoning should be followed for nreaningfirl
research.

Good research also irnplies obtaining reliable data which provides sound
validityto the research findings.
The following principles underlie a good research criteria:
o The aim and objective of the research being conducted should be clearly
specified;
. The researchprocedure should be replicable so that ifthe research needs
to be continued or repeated, it can be done easily;
o The research design should be so chosenthat the results are as objective as
possible.

Self-lnstmctional Material 13
Meaning of Research Interpretation ofanyresearch shouldbe done keeping inmind the flaws in
the procedural design and the extent to which it has an effect on the results.

Research should be calTied out systenlaiically. It should progress in pre-

defined stages, and researchers should avoid using their intuition or
NOTES
guesswork to arrive at conclusions.
Research should be logical so that it is meaningful, and help in decision
rnai<ing.
o Research shouldbe empirical as far as possible.
o The results of the research should only be rised and generalized for .he
population for which the data provides an adequate basis.
o The validity and reliability of the data used in research should be double
checked.
o Further, good research produces results that are examinable by peers,
methodologies that canbe replicated, and knowledge that canbe applied
to real-world situations.

I.11 PROBLEMS ENCOUNTERED BY INDIAN

RESEARCHERS

There are some comrnon problems faced by researchers in developing countries

and India is no exception. Basically, there is a dearth of the tools required for
good research. Many ofthe universities and research institutions are now providing
conputers with Intemet connection to researchers but the facilities provided are
not adequate. Luckily, the costs of both hardware and Internet bandrn'idth are
coming down. While Indian researchers have norv easy access to these tools,
there is still the problem of low visibility of papers published by them. Indian
researchers are often de-motivated to do further research. Other factors like
lack of scientific training in the methodology ofresearch and a non-existent code
of conduct also serve as challenges for the lndian researcher. There is also
insufficient interaction between researchers and the end-users. End-users of
research are the ones who stand to benefit from research and ifthey are not
made aware of the benefit they can gain, getting sponsors to provide funds for
research would be difficult.
There is also a lack ofsafeguards against anyviolation of confidentiality in
data collection. Research studies that overlap one another are undertaken often
and this leads to unnecessary repetition. There is an absence ofresearch culture in
ourcountry.
Other problems that Indian researchers face that are cortmon to developing
countries are:
o Limited or no access to internationalresearchjournals '

o Lack of infrastructure except in a few metropolitan cities

14 SelflnslructionalMaterial
e Low investment in research due to financial constraints
 o lnadequate library facilities and where such facilities exist, they are not Meaning of Research

easily accessible
o Poor encouragement to do research
These problems need to be surmounted effectively in order to promote NOTES
research as a professional activity.

I.I2 MEASUREMENT SCALES

There are three important properties that make one scale ofmeasurement different
from anotlrcr: (i) Magnitude, (ii) Equal intervals, and (iii) Absolute zero.
(i) Magnitude refers to the property of 'moreness'. Any attribute that is being
measured can be more, less or equal at one instance as courpared to another
instance. For example, Vijay is taller than Karan.
(ii) Equal intervals: This means the difference between anytwo points atany
place on the scale has the same meaning as the difference between two
other points that differ by the same number of scale units. For example, the
difference betrveen 20 kg and 30 kg on a weight measurement is the same
as between 80 kg and 90 kg, that is, 10 kg.
(iii) Absolute zero: It is a conditionwhen nothing ofthe propertybeing measured,
exists. For example, ifthere is no pulse measured the situation is alarming,
for the body.
For manypsychological tests such a condition does not exist. For example,
in case of anxiety, there cannot be a 'zero'for there can be no situation where
there is absolutelyno anxiety.
Anymeasurement is a yardstick for evaluation. It involves some formof
judgment. Measurement canbe made ofphysicalobjects, people, situations, etc.
It can be quantitative like height, area, etc. or qualitative like personality traits or
abstract (like patriotism). Measurement is a process of assigning some value to
the observed phenomena. There are clearly accepted rules under which these
evaluations are to be made. Some tasks are easywhile others are complex and
dfficult. Some are direct, like weight. while others like motivation, leadership etc.,
have to be inferred. Some measurements are very precise whereas others are
abstract andtentative. The attempt ofallmeasurement is to achieve confidence in
the evaluations that are made.
Measurements are generallypresented in the fonn of a scale in a range. For
example, ifthe movie is watched by men, women and children, we assign a 0 to
men 1 to women and 2 to children in order to tabulate our observations ofthe
audience. Similarlythe people who attend alone, or as a couple or in a group can
be allotted alphabets like A, C or G for classification purposes. This is the process
ofartificially determining categories. These can fuither be divided qualitatively or
descriptively. There are four levels or tlpes ofscales for measurements-nominal
scale, ordinal scale, interval scale and ratio scale.

Selflnstructionol Moterial l5
 Meaning of Research
Nominal scales: This is a system ofmeasurement where a number is given
to label an event. These are merely convenient labels and do not have any special
meaning or significance. The numbers assigned do not have anyquantitative value.
They do not indicate any order or distance. For example, one player of a baseball
NOTES ,
team is givenjerseyno. 3 and another 5. This is neither a rank, nor a distance
I
between them. It is only a way of categorizing them as a team member. Nominal
scale Measures offer a Count ofData. This is a useful way ofclassification.

Ordinal scales: The ordinal scale places events in an order. It is a ranking

method. For example, Ram stood first in the class and Sita came second. Bharat
was third. However this scale does not make the intervals equal. The real dift',: oflcss
between Ram, Sita and Bharat may not be equal despite their being adjacent in
rankings. This is a 'greater than, lesser than, nrcasurement.
Interval scale: Here, the intervals are adjusted to be equal in terms of
some established rule. For example, the Falrenheit scale. Ifthe zero '0' is accepted
as the startingpoint ofthe scale (water freezes at this point) then measurement
can
be made fromthis absolute zero point. So when temperature rises from 30 to 40
it has the same increment as from 60 to 70. But this does not mean that {0
temperature is twice as wann as 30. This is because both the numbers are based
on the value ofO. The interval of 30-40 and 6G-70 is the same but the ratio of30
and 60 meansnothing because '0'is anarbitrarypoint. Also inpsychologicaltests
the difference in IQ between 70 and 80 is not the same betweenl00 and 110.
Here the l0 points do not have the same meanilrg.
Ratio scale: Here there is an 'absolute zero' formeasurement. For exanple,
the '0'ona centimeter scale, '0' time, '0' errors in a tlped page, etc. The most
precise ofthe scales is the ratio scale because it is easy to make comparisons.
This
makes the scale fairly powerful. IQs are also presented in a ratio forn! to facilitate
comparisons. The nominal scale is the least precise while the ratio scale is the most
precise, with the other two scales falling in between. Behavioral scientists
have to
use the interval scale because ofthe nature oftheir discipline.

However, most behavioural disciprines widelyuse the interval scale.

1.13 SOURCES OF ERROR IN MEASUREMENT

All measurement should be error free in any research, ideally. Since this is not
always possible, it is important to be aware ofsome ofthe sources oferrors:
o Subj ect or respondent factors: Many respondents may be unwilling to
participate in a study. So, ifthey are persuaded to give their responses, they
could be negatively disposed or be indifferent in their replies. This couli
have a significant bearing on the results. Also, a fatigued, anxious or bored
subject maynot provide the true-to-his-heart responses. Thus the subject
could contnbute to the source oferror in the data collected.

16 Ser-Instructional Material
 o Situation: This involves the context inwhich anymeasurement is made. If Meaning of Research
the context places any type ofstress on the individual, then the response
would reflect it. Such a response, therefore, would not be a true index of
the topic ofmeasurement. Even the mere presence ofanother person can
affect the responses obtained. It is important to keep the situation balanced NOTES
and free ofdistorting influences.
o The interviewer / measurer effect: The way questions are put forth can
reflect the bias ofthe experimenter/ researcher. The style, mannerisms,
appearance, language, expressiorl etc., all could distort the presentation.
Any casuakress incarrying out the rneaswement couldhave large inplications
in terms ofthe data obtained.
o Instrumenu tools: Malirnctioning ofthe equipment could instantly lead to
errors inthe measures obtained. Iflanguage is an important component of
the measurenrent, then anrbiguous words, inadequate expressions, inconplete
questions, etc., could all lead to measurement errors. All these factors
must be borne in mind, and carefullyneutralized or completely eliminated as
for as possible to get accurate information.

1.14 TEST OF SOUND MEASUREMENT

Every good measuring tool must be subjected to the tests ofvalidity, reliability and
practicality.
Validity is the most important ofthese three criteria. It means that the test
must measure what it purports to measure. For exanple, ifa test is a measure of
speed, then speed is what that test should measure.

Reliability implies the consistency with which a test measures what it seeks
to measure.
Practicalityrefers to the costs involved in administering a test, the time
needed, the convenience and the ease ofcarrying out the measurements as well
as
the usefulness ofthe obtained data, besides the interpretation.

l.l4.l Test of Validity

A sinple way to determine validity is to ask the question: Is one measuring what is
being thought that the test is measuring?Another wayto determine the valiOity of
a test is the accuracy with which specific predictions can be made from
the iest
scores. This is determined bycomparing the scores obtained with some extemal
test scores as standard. This means one can zusess the individual's present status
or predict his / her future status with respect to some type offunctioning. This
comparison with other relevant evidence is one ofthe better ways ofestablishing
the validity of an instmnrent.

Self-Instructional Moterial l7
Meaning of Research Types of validity:
(i) Content validity: This irnpliesthe extent to whichthe measuring instnrment
has covered adequately all the aspects ofthe topic that is to be measured.
If the measure includes a representative sample ofthe population, then the
NOTES
content validity would be assumed to be good. This can be determined by
a panel ofjudges who evaluate the contents ofan instntment for yielding the
measures the tool is intended to measure. Another way ofestablishing the
content validity is intuitive judgments involving the theme ofthe measuring
tool.
(ii) Criterion-based Validity: Here the success ofthe measuring instrument is
determined bythe ability ofthe scores obtained to predict some outcomes
of a current condition. For example, those who train hard can be predicted
to be winners. The extent oftraining can thus be reflected in the number of
winners.
A good research criterion must posses the following characteristics:
a. Relevance: If the criterion is judged to be the proper measure
b. Freedom from bias: This is reached if everytest taker has an opportunity
to score well. (It is not biased in favour ofany group)
c. Reliabili(v: The measure is stable across several administrations.

d. Availability: The information specified bythe criterionmust be accessible

and available.

i. Predictive validity: The test must be useful in predicting some future

performance from present measures. This is known as predictive validity.
ii. Construct validity: This involves finding out ifthe individualpossesses
a trait or a characteristic. The degree ofthe existence ofthis trait is also of interest
to the assessor. This is the construct on which a test rests; Example: Anxiety,
Intelligence etc. So, the instrument is vaiidated against the construct. The greater
the ability ofthe instrument measuring the construct, the higher willbe the validity
ofthe tool.
iii. Pragmatic validity: This refers to the usefukress ofan instrument as an
indicator or as predictor of some behaviour in the future. So, the validity of a
measure rises from this ability of the tool. Under this approach to validity, two
sub-types are included:
o Concurrent validity: This seeks to see how closely the scores relate to
other known measures of validity.
o Predictive validity: Here the object ofthe scores is to predict the future
situation/ behaviours on the basis ofthe scores obtained fromthe tool
at the present.

Ifthese criteria are met by a tool used for measurement, then the instrument
is thought to be valid. The measures obtained from such a tool can be viewed as a
correct estimate ofthe feature under assessment.
l8 Self-Instntctional Matuial
1.14.2 Test of Reliability Meaning of Research

The test ofreliability of a measuring device is its abitity to yield consistent results
from one set ofmeasures to another.

A valid instrument is always reliable NOTES

Howeveq a reliable instrument is not always a vaiid one. For exarrplc, a machine
may consistentlyunder-weigh by 2 kg. It is a reliable machine, but it is not valid as
a good measure ofweight.

Ifan instrument
is reliable, then temporary and extraneous factors would
not affect the measures obtained.
There are two aspects to reliabitity.
(i) Stability: The extent to which consistent results are obtained with
repeated measurements withthe same instnrment onthe same individual.
A measure ofstability is obtained by comparing the results ofrepeated
measurements.
(ii) Equivalence: This is estimated by comparing the measures obtained
bytwo assessors onthe same aspectisituation or individual.
Reliability can be enhanced by three procedures
(i) By standardizing the conditions under which measurements are made.
Here, all extraneous factors can be kept under control.
(if By systematrzngthe directions for measurement.
(m) Bv training personnel suitably, e.g., technicians who are trained for
measuring blood pressure. Also, having larger samples from the person
on whom the measurement is done.

Types of reliability
There are tlree cofltmon methods ofestimating reliability:
(i) Retestreliability
(i| Intemal consistency reliability
(ii) Parallel forms / or altemate forrns / or equivalent forms. Here, a single form
ofthe test is administered twice on the same sample with a reasonable time
gap.

(i) Retest reliability: Two measures yield independent sets of scores. The
two scores when correlated would give the value ofthe reliability coefficient.
Such a coefficient is also known as the tenrporal stability coefficient. This
means how far the examinees retain their relative position as measured in
terms oftest scores over a given period oftime. The ideal time gap between
thetwo administrations is about l5 days.
(ii) Internal consistency reliability: This indicates the homogeneity ofthe
test. Ifall the items ofa test measure the same function or trait, then the test

Self-lnsttttclional Material 19
Meaning of Research is seento be homogeneous andthe internalconsistencywouldbe high. The
most coflrmon way of determining intemal consistency reliability is by the
split-halfmethod. Here the items to be tested are divided into two equal or
nearly equal halves. Another way to split a test is by using the odd and even
NOTES numbered items. This method is prefened to the regular split-halfmethod
because in a power test, the first half would normallybe made up ofthe
easier items while the second halfwould have the tougher items. The odd
numbers 1, 3, 5, 7,9 and so on and even numbers of 4,6,8, and so on
would balance the items. Each examiner would receive two scores, i.e., the
scores ofthe odd and those ofthe even mrnbers ofa given test. Thus, from
a single administration ofthe same test, two sets of scores are generated. A
'product moment correlation' is conputed to obtain the reliabrlify ofthe half
test. On the basis ofthe reliability ofthis halftest, the reliability ofthe whole
test is estimated. The Spearman/Brown formula is used for estimating the
reliability of the whole test.
This is a useful method two administrations ofthe same test.
as it eliminates
A quick estimate ofreliability is possible. This is a kind of on-the-spot reliability
measure. The demerit ofthis method is that it cannot be used for power tests.
The Kuder-Richardson formula is used for determining the reliability
coefficient in a test where the terms are scored as 0 or 1 or (right or wrong). This
estimates the coefficient alpha which yelds a tneasure of internal consistency.
Alternate forms reliability: These are also known as parallel forms or
equivalent forms or comparable forms reliability. This requires that the test be
developed in trvo forms which are comparable or equivalent. Two forms of the
test are administered to the same sampie either immediately the same day or in a
time interval ofa fortnight. When the reliability is calculated on the basis of data
obtained fromthe two administrations ofthe test, it is the altemate formreliability.
Pearson coeffrcient (r) between two sets of scores obtained from two equivalent
forms becomes the measure of reliability. Such a coefficient is known as the
coefficient of equivalence. Alternate forms reliability measures the consistency of
the examinee's scores between two administrations ofparallel forms of a single
test.

The biggest problemwith this procedure is making the two forms ofa test,
trutyequivalent.
Criteria forjudging whether the forms are parallel:
o The number of Items in both the forms, should be same
o The item-difficultylevels inboththe forms shouldbe similar
o Mode ofadministration ofboth forms shouldbe the some.
Scorer reliability: This kind ofreliabilityis important in tests ofcreativity,
projective tests, etc. This is the reliability which can be estimated by having a
sample oftest independently scored by two or more examiners. The two sets of

20 Self-Instructional Material
scores obtained by each examiner are completed and the resulting correlation Meaning of Research
coefficient is known as scorer reliability.
Test-Retest reliability, internal consistency reliability, and parallel forms
reliability all express reliability in terms ofthe correlation coefficient.
NOTES
1.14.3 Test of Practicality
The practicalitycharacteristic ofa measuring instrument can be estimated in
terms
ofconvenience, cost effectiveness, ease ofadministration, scoring, interpretation,
etc. These factors have an important bearing on the development ofthe most
suitable measuring instruments needed for research.

1.15 SUMMARY
o Research is done to find the solution to a problem, or to know more about
something, or to know new things. Scientific research involves systematic,
controlled, empirical and critical examination ofa hlpothesis orproposition
about the relations in aphenomenon.
o The irrportant stepsneeded forconducting scientific researchare as follows:
(i) Identifying the problem;
(i) Formulating a hypothesis: once the problem has been identified and
the literature review has been completed; the hlpothesis has to be
formulated;
(iif Identifying the variables to be manipulated:
(a) The independent variable, (b) The dependent variable, and (c) The
extraneous variables;
(iv) Formulating a research desrgn;
(v) carrying out the observations and measurements This involves the
-
utilizing ofthe tools ofthe study, for obtaining data; and
(vi) Sumrnarizingthe results;
(vi) Statistical Tieatment;
(viif Drawing conclusions on the basis of the study and deciding upon
applications, further research, etc.
o There are four different ways ofassigning numerals to the attributes ofany
event:
(i) Nominalscale,
(if Ordinal scale, and
(iif Interval or equal interval scale, which includes the characteristics of
the nominaland ordinal scales ofmeasurement, and
(iv) Ratio scale.

S eU- Instruc tio na I Material 2t

Meaning of Research o Some ofthecommon sources oferrors inmeasurement arise from:
(i) Respondent,
(i) Measurer,
NOTES
(ii| Situatioq and
(iv) Testinstrument.
o Sound measurements require that the criteria of validity, reliability and
practicalitybe met. Types ofvalidityare content validity, criterionrelated
validity and construct validity.
o Reliability refers to the consistency of a score. There are two types of
reliability: (i) Test-Re-test method and (ii) Internal consistency reliability
o Test ofpracticalityrefers to the aspects ofadministering a test in terms of
costs involved, ease of administration and scoring, convenience of
interpretation, etc. Economic andconvenience factors are ofsignificance in
testing.

1.16 KEY TERMS

o Fundamental research: A research which focuses on finding
generalizations and formulating theories.
o AppHed research: A research which aims at finding a solution for an
immediate problem facing a society or a businessiindustrial organization.
o Conceptual research: Aresearch that is concerned with some abstract
theory or idea(s).
o Empirical research: Aresearchthat it relies onlyonreal experiences and
observations.
o Quantitative approach: Inthis approachthe data is inthe formofquantities
which is then subjected to mathematical and statisticai approaches'
o Qualitative approach: Approach that deals with data tirat cannot be strictly
quantified, for example, opinions, tastes, attitudes, etc.

I.I7 ANSWERS TO 'CHECK YOUR PROGRESS'

1. A scientific research seeks answers in a systematic and objective way to

certain questions. lt requires inquiry and insight.
2. Fundamental research focuses on finding generaluations and formulating
theories. It gathers knowledge for use in distant future and is termed pure or
basic or fundamental research. On other hand, applied research aims at
finding a solution for an immediate problem facing a societyor a businessi
industrial organization.

22 Self-lnslructional Material
 3. conceptual research is concerned with some abstract theory or idea(s). Meaning of Research
Empirical research, on the other hand, relies only on real experiences and
observations. It is data-based research and its conclusions canbe verified
by observations or experiments.
NOTES
4. In quantitative approach, the data is inthe formofquantitieswhich is then
subjected to mathematical and statistical approaches.
eualitative approach
deals with datathat cannot be strictly quantified, for example, opinions,
tastes, attitudes. etc.

5. Research isborn out ofthe spirit ofinquiry. Researchinculcates a scientific

temper and logicalthinking. The outcome ofresearch is useful inpolicy-
making, for solving economic and business problems and for promoting
industrial growth. Social scientists look for research-based answers for
societalproblems. Research is the process for stimulating intellectual growth
and building a knowledge base.

6. Research methods are the techniques used tbr conducting a research.

Methodology inctudes not only the methods for doing research but the logic
behind the need to use a particular method. The application of certain
statisticalprocedures also forms a part ofthe methodology. Decisions about
size and. type ofdata required are methodological concems.

7" For a research to be good, it shouldbe systematic, logical, empirical and

replicable.

r.18 QUESTTONS AND EXERCTSES

Short-Answer Questions
L List the diftbrent approaches to scientific research.
2. State the objectives ofresearch.
3. What are the criteria ofgood research?
4. State the problems encountered by researchers in India.
5. State the major measurernent scales.

Long-Answer Questions
1. Describe different types ofresearch.
2. Explain the research process in detail.
3. Discuss the objectives types and approaches to scientific research.
4. Examinc the concepts ofvalidity and reiiability along with the various tests
that are available for ensuring them.

5. List and explain the problems faced byresearchers in India.

Se(-lnstructional Material 23
Meaning of Research
I.I9 FURTHER READING

Kerlinger, F.N. 1973. Foundations of Behaviorol Researcfr. NewYork: Holt

NOTES Reinhart and Winston, Inc.
Kothari, C.R. 1995. Research Methodologlt: Methods and Techniqae.r, New
Delhi: Wishwa Prakashan.
Singh, A.K. 2008. ksts, Measurements and Research Methods in Behavioural
Sciences. New Delhi: Bharati Bhawan.

24 Self-lnstructional Material
 Meaning and Definition of
UNIT 2 MEANING AND DEFINITION a Research Problenr

OF A RESEARCH PROBLEM
NOTES
Structure
2.0 Introduction
2.1 Unit Objectives
2.2 Research Problem and Working Hypothesis
2.3 Sample Selection, Types and Size
2.4 Parumetric and Non-Parametric Tests
2.5 Testing Hypothesis I: Parametric or Standard Testing
2.6 Testing Hypothesis II: Non-Parametric Testing
2.7 Correlation and Regression
2.8 Surnmary
2.9 Key Terms
2.10 Answers to 'Check Your Progress'
2.ll Questions and Exercises
2.12 Further Reading

2,0 INTRODUCTION

In this unit, you will learn how to define a research problem. You will also learn
how to select the problem and define it clearly and specifically. The next step is
formulating a working hypothesis. Aworking hypothesis is a tentativeproposition
that provides the solution, in the researchers view, to the defined problem. The
characteristics ofa good hypothesis are presented in the unit.
The next step is testing the hypothesis. For testing the hypothesis there are
two kinds of tests, namely, parametric and non-parametric tests. This unit will
teach you how to choose the right type of test to test a hlpothesis, depending
upon the data. This unit presents a deiailed examination ofparametric and non-
parametric statisticaltests. The different typesoftestsundereachcategoryhave
been presented and discussed.

2.t UNIT OBJECTIVES

After going through this unit, you will be able to :
o Explain how to define a problem and formulate a hypothesis
o Examine the different parametric and non-parametric tests in detail
o Know how to choose from different types oftests to test a hllpothesis

Sell-lnstructiondl l|ateriol 25
Mectning and Definition of
a Reseorch Probtem 2,2 RESEARCH PROBLEM AND WORKING
HYPOTHESIS

NOTES Clearlyidentiffing and defining aresearchproblemas the first step before beginning
any investigation.

The concept of a research problem: There are a few specifications that

have to be met for a researchproblemto exist:
(i) There must be an entity occupying a given environment and there
should exist some uncontrolled values therein.
(ii) There must exist at least two courses ofaction namelyAl andA2 to
be pursued. A course of action is defined by one or more values ofthe
controlled variables, for example, during festival time, the purchase of
certain types of items is one collrse of action.
(iii) There must be at least two possible outcomes, Ol and 02 due to the
course of action. At least one ofthe outcomes should be ofinterest to
the researcher for his study. Then that becomes the objective of
research.
(iv) The courses of action must have some bearing on the objective.
Further; the person concerned does not know which course ofaction
is the best. Then, a problem can be seen to exist. So, the elements that
go into making a research problem are:
a. There should exist some difficulty or problem.
b. There mustbe some objective.
c. There must exist at least two means to reach the objective.
d. The relative efficiency ofthe two aftematives must be considered
for proceeding with the inquiry
(v)There must be a given environment in which the difficulty exists and
persists.
A research problem is an existing problem to which the researcher wants to
find out the best solution" and the course of action to be taken to attain the objective.
The action has to be executed in a given context or environment. Several factors
may influence the outcomes and these have to be kept in mind, while carrying out
the research.
Selecting the problem: The research problem must be carefully selected
to get dependable results. Some tips for choosing a problem.
o Avoid choosing a topic that has alreadybeen researched, extensively.
o Avoid controversial topics, while starting a research career.
o Costs, tirne. competency, etc., need to be kept in mind.
o Carry out a small pilot study, before launching into the final investigation,
ifpossible.
26 Sell:lnstntctionol lfurrt'riul
 o The boundaries within which a problem to be studied should be clearly Meaning and DeJinition of
a Research Probtem
defined. This is to be followedtyfirst posing a problem for the study
then determining the processes and techniques for the study.

Steps involved in research NOTES

(i) Prepare a general statement ofthe problern to be studied
(ii) Surveyingthe available literature
(iii) Developing ideas through intuition and discussions
(iv) Formulating a working definition
statement of the problem: Begin with a broad field and then narrow it
down to specifics in order to search a limited area in terms of a specific, single
problem.
understand the problem: This is done in terms of its origin background
and other factors.

survey the available literature: This is to become familiar with the

relevant theories in the field of study and all other records ofresearch work that
have been carried out in the field so far. Knowing about the field helps one to
focus on the research topic, bctter.
Developing ideas: Ideas can be developed through one's own intuition,
reading the background literature and through discussions with others in the related
field.
Phrasing the research problem: This means stating the research problem
in operational terms. A statement ofall technical terms should be made clearly.
The basic postulates ofresearch should be identified. A reasonable time period is
to be specified, for the study. The scope ofthe investigation and the limitations
within which the study is to be carried out is to be specified.
Example 2.1: why are goods manufacftred in lndia so much cheaper than those
produced in the West I
This is a general problem. Then it could be examined, discussed and
formulated as a working hlpothesis. Aworking hypothesis is a supposition based
on previous observations that is offered as an explanation for the observed
phenomenon and is used as the basis for further observations, or experiments.
The task of research is to formulate a research problem, study it systematically
and then test the hllpothesis for acceptance or rejection.

Example 2.2: lVhat are the factors responsible for the chinese goods being
cheaper than similar Indian products? Here the focus must be on cost factors and
pricing methods adoptedbythe manufacturers inthe two countries.

To conclude, research has to be carried out by being well informed,

developing hunch about the topic to be studied by extensive reading about related
a
work, discussions and ifpossible carrying out a pilot project. Then, the working

Self-lnstruclional l'lalerial 27
Meaning and DeJinition af definition needs to be fully operationalned,the techniques, procedures identified,
a Research Problem
sample selected and the timeframe, costs, etc., have to be estimated. Then the
research has to begin.
A few conditions about a research question:
NOTES
o It is clear and unambiguous
o [t is fairly specific
o lfidentifies the type ofdatato be collected
o The questions asked are related to each other
o It is aworthwhile attempt
The problem statement of any research is written in the form of a question.

Examples 2.3:
(i) What is the relationship between stress and performance?
(ir) Does leaming proceed beuer in democratic environment or in an authoritarian
structure?
A problem also expresses the relationship between two or more variables.
So, one ofthe variables could be manipulated and its efects on the other, studied.
Example 2.4: Is higher anxiety related to lowered performance?
A problem should be testable by empirical methods. Data obtained should
be testable.

It is wise to avoid ethical and moral statements like for example should gay
people be allowed to join the defiance services. The problern chosen should not
be veryfivial, extremelybroad or generalthat cannot be studied effectively. Similarly
excessive specificity can also be a lirniting factor.
Considerations for selecting a problem: Before a problem for
investigation is selected, several qttestions have to be asked.
(i) Is the problem significant in terms ofthe variables to be studied and its
contribution to research, social theories, practices, etc.
(if Can useful data be gathered for the study?
(O Is there anyduplicationthat couldhappen?
(iv) Is the problem operationally workable?
(v) Are there good instruments to measure?
A problem is worth studying, ifthe answers to most ofthese questions are
positive.
How does a problem show up for study?
Does a gap inthe information exist?

Contradictory results occur when certain facts are unwarranted from the
existing knowledge.

2E Sel.l-lnstnK'tional Maleri.tl
 Research problems are oftwo types Meaning and Delinition of
a Research Problem
(i) Solvable: For solvable problem an answer is possibre. Ahypothesis
canbe formulated.
(i) Unsolvable: Forunsolvable problem, no answers are possible.
NOTES
Example 2.5: Is there life after death?

Meaning and characteristics of a good hypothesis

while conducting research, the first step is to identify the problem. The next step
is to researcher should formulate a hypothesis which is a tentative solution in the
form ofa testable proposition.
This testable proposition is called a hypothesis. Hypothesis is, therefore, a
suggestion about a possible solution. Hypothesis seeks to studythe relationship
between two variables. It is a potential association between variables. It is a
conjecture o f the relationship between two variable s.

Differences between a problem and a hypothesis

Problem:
1. Is not testable
2. Is stated in the form ofinterrogative sentences.
Example 2.6: What is the relationship between reward and learning?
Hypothesis: It is testable and is stated in the form ofa declarative sentence.
Example 2.7: Reward facilitates learning

Criteria of a good research hypothesis

(i) A good hypothesis should be clear and based on operationally defined
concepts.
(if It mustbe testable.
(nD It should be short, clear, specific and simple (in the use ofwords).
(iv) It should relate to the existing body oftheory and fact.
(v) It should display logic and comprehensiveness.
(vi) It should be general in scope.
(vii) It should be in accord with other existing hypotheses in the field of study
Types of hypothesis
o Causal hypothesis
o Descriptive hypothesis
Causal hypothesis explains a tentative influence of one factor on another.
Example 2.8: Frustrationis the cause ofaggression (causal).
Descriptive hypothesis: Merely describes the behaviours.
Se[-Instructional Material 29
Meaning and Delinition of Example 2.9: Frustration frequently could result in aggression.
a Research Problem
Other types of hypothesis are simple hypothesis. complex hypothesis,
research hlpothesis, null hypothesis, and statistical hypothcsis.

NOTES Simple hypothesis: It contains only one or two variahles

Complex hypothesis: It has more than two variables.
Research hypothesis: It is derived fromresearcher's theory. It is called a
working hypothesis
Null hypothesis: The null hlpothesis is in a sense the opposite ofthe research
hypothesis. It is a hypothesis that negates differences. In fact, researchers are
interested in rejecting this hypothesis; so that the reverse, i. e., the research hypothesis
is accepted.

Statistical hypothesis: This is also known as the alterative hypothesis

(//,). Here the null hlpothesis (I/o) and the alternative are expressed in numerical
tenrs.
Example 2.10: On the basis of the existing psycho-social theory one would
predict that two groups would differ on the measure of anxiety. So, the research
hypothesis would predict that the two groups would differ.

To test this hypothesis, it is stated in an operational way; as the alternative

hlpothesis, that is 11,. The operational way to state the alternative hypothesis
would be that the mean anxiety scores of these two groups would differ or are
unequal. The null hlpothesis (Il) would implythat the mean ofthe anxiety scores
ofthe two groups is the same. Ifthe data suggests that Hnis to be rejected then /1,
is true and supports the research hypothesis.

CHscrYouR PRocRESS

l. What is the meaning ofresearch and how does one define a research
problem?
2. How does a researcher select a research problem?
3. How is a research problern stated?
4. What are the different typcs ofresearch problems?
5. What is a hypothesis?
6. What isanullhypothesis?
7 . What is statisticalhypothesis.

30 Self-lnsn'uclional lvtuteriul
 ivleaning and D.fi,.ition o.f
2.3 SAMPLE SELECTION, TYPES AND SIZE a Research Problem

Sample selection
I
NOTES
All research in the field ofbehavioural sciences involves drawing inferences from a
]

specified, identifiable group on the basis ofa selected sarrple. The clearly identifiable
and specified group is known as the population or universe. The selected group of
persons or objects is called the sample. The conclusions are drawn fi'om the sarnple.
which are deemed to be ralid to the entire population. Such conclusions arc known
as the statistical inferences.

Apopulation can be finite or infinite. Afinite population is one where all

the members can be counted. An infmite population is one where allthc members
cannot be counted (e.g., stars in the sky). Apopulation can be irnaginary or real.
Ameasure based upon the entire population is called a parameter. A
measure based upon the sarnple is called a statistic.
Asample is lirnited number or set ofpersons or elements that are chosen
from to some plan. lt is thought to represent the population.
a population. according
Samples are based on probabilities. Probability is a form ofrelative frequency.
For example, the probability ofa seeing a head when coin is tossed once is l/2 or
0.5. Probability is expressed as a fraction or in decimal numbers.

Need for sampling

Sampling is needed for several reasons:
Sampling is economical in terms oftime and money. It saves time as well as
cost because only a limited sample is involved, not the entire population;
Sarrpling is carried out by trained personnel, so it has considerable accuracy
in measurement and testing;

Sampling errors canbe easilydetermined fromthe sampling process. This

yelds valid information about the population chamcteristics; and
Sampling is the only available procedure when the population is infinite;
Sampling enables fairly accurate generalizations about the population from
the studyofthe sample.

Steps in sampling
Sampling is the plan for obtaining a sample from a given population. There are
several steps that a researcher must keep in mind while selecting a sample:
The type ofuniverse or population to be studied can be finite or infinite.
Sampling unit: This stands for a geographical area like a district or province.
taluk, etc., or it could be a social unit like a family, school club, etc. The researcher
has to decide the factors that would be studied in advance. Alist ofall the sarnpling
units is called a sampling frame. This is the source list and contains all the items of
a universe or population. Self-lnstnr:tional Moteriol 31
Meaning and Definilion oJ' Types of sample
a Research hoblem
Most samples can be categorized into two types:
(i) Probabilitysampling
NOTES (i!) Non-probability sampling.
(i) Probability sampling: It sanple is based on the concept ofrandom selection
or chance sampling. Here, every item ofthe universe has an equal chance of
inclusion in the sample. It is a form of lottery method where the units are
chosen fromthe whole group bya mechanical method. This is almost a
blind selection. Random sanpling ensures that the law ofstatistical Regularity
is followed. This implies that ifthe sample chosen is a random one, the
chances are that the sample will have the same composition and
characteristics as the universe. This is why randorn selection is considered
as the nmst useful method for obtaining a representative sanple. Probability
sanpling must follow the conditions given below:
a. The size ofthe parent populationmust be knownto the investigator;
b. Each element ofthe parent populationmust have an equalchance of
being included in the sample;
c. The sample size needed must be clearly specffied.
The major probability sanpling methods are:
a. Sinrple random sampling;
b. Stratified random sampling - (a) Proportionate stratified random
sampling, and (b) Disproportionate stratified random sampling;
c. Area orcluster sampling.
(ii) Non-probability sampling: This is a method in which there is no way of
determining the probabilitythat each iteminthe population gets inchrded in
the sample. This means that there is no basis for estimating how closely the
characteristics ofa sample approximate the parameters ofthe population
from which the sample has been drawn. This is due to the absence ofrandom
selection procedures. The nrajor techniques used in non-probability sanpling
are:
a. Quota sampling
b. Accidental sampling
.. c. Judgmentalorpurposivesampling
d. Systenutb sampling
e. Snowballsampling
f Saturationsampling
g Dense sampling

32 Se$-lrctuctionalMatefial
 Size of the sample Meaning and Definition o.f
a Research Problem
Size refers to the number ofitems selected fromthe universe to constitute a
sanple.
The size ofthe sanple should be neither too large, nor too small. An optimum
sample size should be:
NOTES
a. Efficient
b. Representative
c. Reliable
d. Flexible
The sample size should be decided bythe level ofprecisionneeded and the
estimate ofthe confidence level, desired. The size ofthe population variance is an
important determiner. Ifthe population variance is large, then a larger sample size
is indicated. The size ofthepopulation is another factorto be kept inmind. This
limits the size ofthe sarnple

other aspects to be considered for determining sample size:

(i) The pararneters of interest; 1. Estirnation ofsome characteristics or some
proportion ofpersons in a population; 2. Knowing solne average measure
ofa population; and 3. Knowing about some sub-groups ofa population.
All these types of estimates have a bearing on the sample size.
(rl Budgetary constraint.' - From a purely practical point ofview, the size of
the budget is bound to influence the size ofthe sarnple.
(i\ sampling procedure: the choice of the rnethod in selecting the sample
would have to provide a balance between the cost involved and the least
possible sampling error that would result fromthe sample size.

2.4 PARAMETRIC AND NON.PARAMETRIC

TESTS

These tests are utilized extensively in behavioural research.

A parametric statisticaltest is one in which certain conditions, about the

populationparameter fromwhichthe sample is drawn, are specified.
The conditions forparametric testing are:
o The observations must be independent, i.e., the selection ofone case
must be dependent upon the selection ofanother case.
o The obseruations mwt be drawn from a normalty distributed population
o The samples drawn from a popuration must have equal variances. This
is more so, when the sample size is srnall. This is the condition of
homo geneity o f variance
o The variables must be expressed in interval or ratio scales. Nominal or
ordinal scales do not qualiff for
a parametric test.

Self-Instructional Material 33
Meaning and Delinition of o The variable studied should be continuous. Examples ofparamedic tests
a Research Problem
are z test, / test and Ftest.
A non-parametric test is one which does not speciryanyconditions about
the parameters ofthe population fromwhichthe sample is drawn. These are also
NOTES
called distribution free statistics because no assumptions are made about the
distnbution ofthe population. For these tests also, the variables must be continuous
and the observations should be independent. But these conditions do not apply
rigidty.
Examples ofnon-parametric tests are the chi-square test, the Mann-Whitney
Utest, Kendall's I, Kendall's coefficient of concordance W,etc.
Conditions that suggest the use of the non-parametric test:
o When the shape ofthe distribution ofthe population is not a normal one.

r When the variables have been expressed as nominal measures or frequency

counts
o When the variable have been expressed as ordinal measure (ranking)
These tests are less precise and therefore run the risk ofnot rejecting the
nullhypothesis, whenit is false.
Advantages of non-parametric tests :
o These tests allow the statistics to be computed from simple formulae.
o Greater applicability
o Calculations faster, specificallywhen the size ofsample is small
o Assumptions being easier, less violations occur.
o Nominal and ordinal scales measures are acceptable
o These tests are useful for mall samples. But as the size increases, the statistics
become cumbersome.
o Non-parametric tests are more efficient andconvenient, specificallywhen
the assumptions ofparamedic tests are not met.
Disadvantages: When the sample size is large (above 30) they give lower
statistical efficiency. Ifthe assumptions ofparametric statistics are met, the use of
non-paralnetric tests means a waste ofuseful data. [t is not easy to find tables for
testing significance ofdata.

Cnucx YouR PRoGRESS

8. What is aparametric statistical test?

9. What is a non-parametric statistical test.

34 SelfJnstructional Materiul
 Meaning and Definilion of
2.5 TESTING HYPOTHESIS I: PARAMETRIC OR a Reseqrch Problem

STANDARD TESTING

The most important parametric statistics are: NOTES

(i) Student's / test and e test
(ii) FRatio
(iii) Pearson r'

(i) Student's , test and z test

A rescarcher can use cither the / test (or I ratio) or z test (or z ratio) to test the
siguificancc ofdillerence bctween two mcans. The computation ofr or; involves
computing the ratio between the experimental variances (that is. the difference
between two means) and the error variance (that is, the standard error ofthe
mean difference). However, there is one basic difference between t ratio andz
ratio. If the sample size is less than 30, one uses the I test or student's I for testing
thc signiticance ofthe difference betweenthe two means. This small sample size
test was developed in 19 I 5 by William Seely Gosset, a statistician. This statistic is
known as student's r. When the sample size is more than 30, the ratio of the
difference between the two sample means to the standard error of this difference
is calculated by the ; ratio which is interpreted through the use ofnorrnal probability
tables. However, one should note that as the sample size increases, the critical
,,alues oft necessary for rejecting the null hlpothesis graduallyreduce and approach
thc : values ofthe normat probability table.

Before discussing the small sample / test in detail, five important concepts,
namely. degree offreedom, null hypothesis, level of significance and one-tailed
test should be understood.

Degree of freedom
The degree of lieedom means freedom to vary and its abbreviated expression is
d.f
In statistical language, the degree of freedom means the number of
observations that are independent ofeach other and that cannot be deduced from
each other. Suppose we have five scores and the mean of five scores is 10. The
fifth score immediately makes adjustrnent with the rernaining four scores in a way
which assures that the mean of all five scores must be 10. For example. suppose
we have four scores 12, 18, 5, 12, and for the mean to become 10. the fifth score
must be 3. In another distribution, ifthe four scores are2,10,8, 5, the ffih score
must be 25 in order to derive a mean of 10.
The meaning ofthis is that four scores in the distribution are independent,
they may have any values and they cannot be deduced from each other. The sizc
ofthe fifth score, however, is fixed because the mean in each case is 10. Hence
d.f:N-l:5-l:4.
Self-lnstructional Moteriol 35
Meaning and Delinition ol' In larger cases, in computing the mean it goes something like this: Suppose
a Research Prohlem
we have a set of I 0 1 scores. We compute the mean and in computing the mean,
we lose 1 d.f . We had initially 101 d.f (because there were 101 scores) but now
after computing mean. we have.ly'- I : 101 - I : 100 degrees of freedom.
NOTES
Sometimes, we have paired data and in such cases, the number of degrees of
freedom is equal to one less than number ofpairs.

Null hypothesis
The starting point in all statistical tests is the statement ofnull hypothesis (F/n),
which is a 'no difference hypothesis.' A null hypothesis states that there is no
significant difference between the samples under study. It makes ajudgment pbout
whether the obtained difference between the samples is due to some true differences
or due to some chance error. The null hypothesis is formulated for the express
purpose ofbeing rejected because ifit is rejected the altemative hypothesis (11,)
which is an operational statement of the investigators' research hypothesis, is
accepted. Aresearch hypothesis is nothing but predictions or deductions drawn
from a theory. The tests of the null hypothesis are generally called tests of
significance, the outcome of which is stated in terms ofprobability figures or levels
ofsignificance.
Ifthe difference between the experimental group and the control group is
very sma[ the experimenter is likelyto accept the nullhypothesis, indicating the
fact that the small difference between these two groups is due to sampling errors
or some other chance fluctuations. On the other hand, ifthe difference between
the experimental group and the control group is too large, the experimenter is
likely to refuse or reject the null hypothesis, including the fact that the obtained
differences are realdifferences between or among the samplesunder study.

Level of significance
The null fupothesis has been developed for the express purpose ofrejection. The
rejection or acceptance ofthe null hlpothesis is based upon the level ofsignificance,
which is used as a criterion. The levels of significance are also known as alpha
levels.

In psychological, sociological and educational researches, there are two

levels ofsignificance which are commonlyused for testing the null hypothesis.
One is the 0.05 or 5 per cent level and another is the 0.01 or I per cent
level ofsignificance. Ifthe null hypothesis is rejected at the 0.05 level it means that
5 times in 100 replications of the experiment the null hypothesis is true and was
false in 95 other replications, which means that there is 95 per cent probability of
the obtained results are due to experimental treatment rather than due to some
chance factors. Rejection of the null hypothesis when that hlpothesis is true
constitutes a Type I error or alpha error. Thus it canbe said that at the 0.05 level
ofsignificance the experimenter commits a 5 per cent type I errors when he rejects
36 Selflnstructional Material
a null hypothesis. Some investigators may want a more stringent test and the 0.01 Meaning and DeJinition of
a Research Problem
level is one such level where the investigator commits a Type I error I per cent
only. The 0.01 level suggest that a99 per cent probability exists that the obtainecl
results are due to the experimental treatment, and hence, once in 100 replications
ofthe experiment, the nullhypothesis wouldbe true. Anevenmore a stringent test NOTES
of significance is the 0.001 level, which is not commonly used in behavioural
researches. This level suggests that onlyonce in 1000 replications ofthe experiment,
the null hypothesis would be a true one and in 999 replications (out of I 000) the
obtained results can be attributed to the experimental treatment. In testing the
significance ofthe statistics, sometimes the investigator accepts the nullhypothesis
when, in fact, it is false. This error is technically known as the Type II error or beta
elTor.

From the above discussion, it can be said that an error ofrype I can be
reduced byputting the alpha level at the 0.01 or 0.001 level. But as one reduces
the chance for making a Type I error, one increases the chance for making a Type
II error where one does not reject the null hypothesis when it shouldbe rejected.
Therefore, while decreasing the possibility of making one type of error, the
investigator also increases the probability ofmaking another type oferror. The
research workers must be cautious with this situation and should try as far as
possible, to limit the probability for making a Type I error.

One-tailed test and two-tailed test

One+ailed test is a diectionaltest, which indicates the direction ofthe difference
between the samples under study. Suppose the experimenter conducts an
experiment inwhich one takes two groups--one is the controlgroup and another
is the experimental group.
only the experimental group is given training for 5 days
onvarious kinds ofarithmetic operations. Subsequently, an arithmeticalabilitytest
is administered on the two groups and the scores are obtained. In the above
situation, there is reasorr to say that the mean arithmetical score ofthe experirrental
group will be higher than the mean arithmetical score ofthe control group. This is
the alternative hypothesis, which indicates the direction ofdifference. When the
altemative hypothesis states the direction ofdifference, it constitutes a one-tailed
test. The null hlzpothesis would be that the mean ofthe experimental group is equal
(no difference) to the mean ofthe control group. Ifthis is rejected we accept the
above alternative hypothesis.

The above facts canbe schematicallypresented as follows:

d. Ho: Mr: Mr(no difference between Mrand Mr)
Alternative hypothesis :

b' Hr: Mr* M,

c. Hr: Mra M,
d' H,: M,'M,

Self-lnstrilctional Materiql 37
Meaning and Definition of where M, : meanof the experimental group and
a Research Problem
M r: mean of contro I group.
When it is said that the mean ofthe experimental group willbe higher than
the meanofthe controlgoup, one is concemedwithonlyone endofthedistribution.
NOTES
Putting it in terms of a normal curve, one is concemed with only one end ofthe
curve (see figurebelow). Whenthe alpha level is set at the 0.05 level, a 5 percent
ofarea ofthe normal curve is obtained, all in one-tail rather than having distributed
it equally into two tails ofthe curve. Therefore, the directional null hlpothesis is
called a one-tailed test. A simple inspection of the table of areas of the normal
cr:rve givenat the end reveals that a z score of 1.64 cuts off5 per cent ofthe area
under the normalcurve inthe smaller part, and similarlyaz score of2.33 cuts off
I per cent ofthe area in the smaller part. Ifthe null hypothesis is rejected, that is
hlpothesis I is not tenable, we automatically accept the alternative hypothesis. If
the experirnenter has some reason to believe that the experimental group would
have a lower mean score thanthe control group (altemative hypothesis), he can
set up a directional hypothesis that the mean of the experimentalgroup is lower
than the mean of the control group (one-tailed test). Rejection of this hypothesis
would automatically lead to the acceptance ofthe null hypothesis. This time the
normal curve in which the experimenter is interested is 5 per cent or 1 per cent of
the area in the left-hand tail of the normal curve. When the null hypothesis is
rejected by using a one-tailed test, one must say that one is rejecting the hypothesis
at I per cent or 5 per cent points, not levels (see Fig. 2.1).

Fig. 2.1 One-Tailed kst al 0.05 or 5l'/o Poittts

Source: Singh, A.K., Tests, Measurements and Research Methods in behavioural sciences,
2008.

A two-tailed test is one in which the investigator is interested in evaluating

the differencebetweenthe groups. The direction ofdifference is ofno importance
here. The nullhypothesis willbe that where the mean ofthe experimental group is
equal to the mean ofthe control gouP, that is, there is no difference between the
means ofthe experimental group and the control group. The ahernative hypothesis
would be that the mean ofthe experimental group is not equal to the mean ofthe
control group. Thus, the investigator's concern is shown with both tails of the
distribution. At the 5 per cent level, there is 5 per cent ofthe area under the normal
curve equally divided at both the tails (see figure below). From the table of areas
ofthe normal curye, it can easilybe read thataz score of+l.96 cuts off2.5 per
3E Self-lnstructional Material
cent area at both extremes. Since a normal curve is bilaterally symmetrical, a z Meaning and Definition of
a Research Problem
score of-I.96 would also cut off2.5 per cent area in the left-hand tail. In a two-
tailed test a negative z score is interpreted in the same way as a positive z score.
When the null hlpothesis is rejected by using a two-tailed test, it is said that it has
been rejected at 5 per cent or I per cent levels and not at 5 per cent or I per cent NOTES
points (see Fig. 2.2).

Fig. 2.2 Tv,o-Tailed Test at 0.05 or 5o,4Level

Source: Singh, A.K., Tests, Measurements and Research Methods in behavioural sciences,
2008.

The I ratio calculation from different groups

Ordinarily, three types ofsituations exist while one is calculating the r ratio:
a. tratio fromindependent groups
b. lratio fromcorrelated groups
c. I ratio from matched groups
Regardless ofthe nature ofthe Broup, / ratio is calculated bythe following
equation:

(M,- Mr)-0
,l:- SE,
...(2.1)

where M, :meanofthe first group

M, : meanofthe second group
SD, = standard error ofthe difference betweentwo sanple means.
In each ofthe above three groups, the formula for SE varies and hence, the
following separate calculation ofthe t ratio fromeach ofthe above types:
t ratio from independent groups: Two groups are said to be independent
whenno correlationexists betweenthem. Suppose one group ofboys (N= 20)
and one goup ofgirls (N:22) were administered a mechanical reasoning test.
Their data are summarized as:
N, :20 Nr:22
M, :34.56 Mr:30.56
SD, :5.68 SD, :6.98
Self-lnstructional Material 39
Meaning and Definition of Do the two groups differ onthe measure ofmechanicalreasoning test?
a Research Problem
The above question can be answered by computing the / ratio. For
independent groups, SE may be calculated with the help ofthe following question.

NOTES SEo: +sE i't; ...(2.2)

where SE, : Standard error ofthe difference between means;
5E r, : Standard error ofthe first mean; and
5E ,r: Standard error ofthe second mean.

Here standard error ofmean can be calculated as given below:

SD
SE, = ---- ...(2.3)
VM_ I

Substituting the value in Equation (2.3), we have

sEr,, =
5.68
,E-=1.302;
g0u, =# =1.s24

Now, ^SErbecomes equal to:

SE
o = .i1t.:OZy' + (1.524): = 2.004
(34.s6-30.s6)-0 4.00
= 1.996
2.004 2.044

d.f = (N, -l)+(N, -l) = (20 -l)+ (22-l) = 40

Entering the probability table of t ratios we find that the obtained r at
d.f = 40 is significant at the 0.05 level but not at the 0.01 level. Hence, we conclude
that the null hlpothesis is rejected and that there is a true difference between the
meahs ofthe groups ofboys and girls.
r ratio from correlated groups: The correlated groups are those which
exhibit some correlation with each other on the given measures. One ofthe most
convenient ways ofgetting the correlated means is to repeat the same test on the
same groups twice on fwo different occasions. In between the initial and the final
administration of the test, some experimental treatment is given to the group.
Because the group and the test are the same, it is highlyprobable that there willbe
a correlation between the initial measures and the final measures. Suppose a group
of20 students ofClass IV is administered an English spelling test in January. The
obtained mean score and standard deviations are given below. After a year's
training in spelling, they are again given the same test. This time their mean is
considerably raised and the standard deviation is lowered. The correlation
coefficient between the initial and final set of scores was positive and significant.
Does training produce a significant difference between the initial mean and
the finalmean?
40 Self-lnstmctional Material
 Final test Meaning and De.finition of
a Research Problem
Ns
Means 36.28 40.33
SDs 4.28 3.32
Coefficient of correiation 0.80 NOTES
In case oftesting the significance ofmean difference between two correlated
means, SEo is calculated with help ofthe following formula:

SE, : ...(2.4)
where r,, : coefficient ofcorrelation between the initial set ofscores and the final
set of scores. The rest of the subscripts are defined like those in the equation
above.

The standard error ofmean (sE) is calculated by the equation below:

3Er,: 5Er,:
#=0.982 #=0.s32
SE, :
:
:m=0.641
(M,- Mr)-0 (36.38-40.33)-0 _ 4.05
" SE, 0.625 0.641
= 6.31

df :N- l:20-l: t9
Entering the probabilitytable ofl :
ratios at d.f 19, we find that our obtained
value of I exceeds the value of r at even the 0.001 level ofsignificance. Hence, the
null hypothesis is rejected and it is concluded that the training has produced
significant difference between the mean ofthe initial set ofscores and the mean of
the final set ofscores.
t ratio from matched groups: Sometimes it becomes necessary for the
researchers to match the groups under study. Matching can be done on the basis
ofnumbers or it can be done in terms ofmean and standard deviation. When the
matching is done on the basis ofthe number ofthe subjects, each person has his
corresponding match in the other group and therefore, the number ofpersons in
the two matchedgroups is always equal. whenmatching is done interms ofmean
and standard deviation, the number in the two groups may or may not be equal.
Suppose two groups fromtwo different classes are considered and each
group is compared on the basis ofnumerical reasoning test. Both groups have
been matched in terms of mean and standard deviations on the basis of their
scores on a General Intelligence Test. Do the groups differ in terms of mean
numerical ability? The following is the data obtained:

Self-lnstuctional Material 4l
Meaning and Definition of Class VIII Class X
a Research Problem
NI 100 r20
Means of Intelligence tests 70.26 70.25

NOTES SDs on Intelligence Tests 9.98 10.02

Means on Numerical Reasoning Tests 55.62 60.34

SDs on Numerical Reasoning Tests 8.67 7.58

Coeffrcient of Correlation between

General Intelligence Test Scores and numerical reasoning test scores 0.45

The equation for calculating sE, in case ofgroups matched in terms of

means andSD is givenbelow:

sEo:J(*", +sEM:){r-r') ...(2s)

where subscripts are defined as usual. Now, we can proceed as under:

8'67 7E
tfrro
= o'867; 5Eu, = 0.691
5E r,.: ffi
SEo: -
:
(M,- Mr)-o _ (ss.62-60.34)-0
,, _ SEO 0.991
= 0.991

4.72
= 47.62
0.991
df : ({ - 1) + (Nr- 1) -1 : (100- 1) + (120- 1) - 1

: (99 + 119) - | : 218 - | :217

Entering the probabilitytable ofr ratios at d.f 217 ,we find by interpolation
that the obtained value of / exceeds the / value required at even the 0.001 level of
significance. Rejectingthenullhypothesis, weconclude that the two groups differed
significantly in terms ofnumerical mean ability.

(ii) F ratio
The /ratio or zratiois one ofthepowerfulparametric tests th,roughwhichwe can
test the significance ofthe differencebetween two means. There afe two general
limitations ofthe r ratio. First, when there are several groups and ifwe want to test
the significance ofthe mean difference among thenr, several I ratios me required to
be computed. For example, suppose there are five groups. Then there is need to
compute.

42 Self-lnstructional Material
 N(1/ - l) 5x4
l0r ratios
Meaning and Definilion of
a Research Problem
)', ,,,(2.6)
Then it becomes a cumbersome task. Secondly, the / ratio does not account
for interaction effect in its statistical analysis. The variations in the scores may be
NOTES
due to the interactions taking place among groups. Such variations are not accounted
for by I ratios. In order to remove these two limitations we turn to analysis of
variance, originally developed by R.F. Fisher. Analysis of variance is a class of
statistical techniques through which we test the overall difference among the two
or more than two (normally more than two) sample means.
Analysis ofvariance is of two common types: Sinple analysis ofvariance
or one-way analysis ofvariance and complex analysis of variance or two-way
aualysis ofvariance. Analysis ofvariance (ofwhatever types) is oftenreferred to
by its acronyn, ANOVA.
In simple analysis ofvariance there is only one independent variable and the
samples are classified into several groups on the basis ofthis variable. Since the
basis ofclassification is onlyone independent variable, the sinple analysis ofvariance
is also known as one-wayANovA. SuchANovA is suited to the completely
randomized design. In complex ANovA, there are two or more than two
independent variables, which form the basis of classification of groups. Such
ANOVA is suited to factorial design.
Statistically, the Fratio is calculated as follows.
Larger variance Between-qrouDs variance
F* Smaller variance
- nfvr

Within-groups variance ...(2.7)

Between-groups variance refers to variation in the mean of each group

from the total or grand mean ofall groups. within-groups variance refers to the
average variability ofscores within each group. The theme ofthe analysis ofvariance
is that ifthe groups have been randomly selected from the populatiorl these two
variances, namely, between-groups variance and within-group variance are the
unbiased estimates ofsame population variance. The sigruficance ofdifference
between these two types ofvariances is tested through the Ftest.
The following table illusrates the sinple anallsis ofvariance, with independent
measures. One can see that the scores ofthe three groups ofsubjects A, B andC
are shown on the educational achievement test, and ANOVA has been calculated
fromthose scores.
Grand sum (XX): 561 + 154+ 312:1027
Grand sum of squares (X.If): 386 I I + 2610 + I 6070 : 57 29 l

Step I : Correction (q'.Ngp Qo2T'z

=35157.63
30
Step 2: Total sumofsquares (7SS): ZX -C
: 57291 - 35157 .63 :2213337
Self-Instruct onaI Material
Meaning ond Definition of Table 2.1 Simple ANOVA based on the Scores of Three Groups
a Research Problem

Gr. A Gr. B Gr. C Gr. A Gr. B Gr. C

X x X x r x,
NOTES 78 l0 50 6084 r00 2500
58 t4 58 3364 t96 3364
40 9 65 1600 8l 4225
30 t2 72 900 144 5184
l0 20 l0 100 400 100
65 25 9 4225 625 81

88 t2 8 7744 IM &
87 18 r0 7569 324 100
80 t4 t6 6400 196 256
25 20 t4 625 400 t96
Sums 561 t54 312 3861 I 2610 16070

Step 3: Between (or among) sum of squares (BSS):

(24)' +\>x) ...+ Gx.)' n,_c

nt n2 nn

_ (561)'z
+(154)'z +Q1^2)z _35157.63 ...(2.8)
l0 10 10

314721+ 237T6 +97344

-35157.63
l0
: 43578.10 - 35157 .63 : 8420.47
Step 4: Within sum of squares: 7SS - B,SS : 22133.37 - 8420.47
:13712.90
Tsble 2.2 Summary: Analysis of Variance

Sources of variation d.f. Sum of squares Mean squares or

vartance

Between-groups K-l:3-l:2 8420.47 4210.235

Within-groups N- K: 30 - 3 :27 13712.90 507.885
Total N- I : 30 - l:29 22133.37

Between-groups variance _ 4210.235 =g.2g9 =g.29

r : within€roups variance 507.885

Since there are 30 cases in the above problem we have I 30 I N- : -

: 29 d.f. in all. d. f for groups (between-groups) is equal to the number or groups
(K) minus one. Since there are three groups d.f. for between-groups is K- 1 :
44 SeU'-lnslructional Malerial
 - :
3 1 2. d.f. for within-groups is equal to the total nurnber of cases minus Meaning and Definilion of
:
number of groups (K). Hence, it is equal to N - K 30 - 3 : 27 . Aft.er putting the
a Research Problem

number ofdegrees offreedorrl the sum ofsquares for each ofthe three sources of
variations, we compute mean squares or variances, which are obtained by dividing
each ofthe sum of squares by its respective number of degrees of freedom of NOTES
freedom These two types ofvariances are the estimates ofthe population variance.
we obtain F ratio by dividing the between-groups variance by the within-groups
variance.
F ratio is interpreted by the use of the Ftable (Guilford & Fruchter, 1 97 8).
In the F table the number ofdegrees offreedom for greater mean square (d.f is
r)
written at the top and the numberofdegrees for freedomforsmallermean sqrnre
(d.f.r) is written on the left-hand side. For this problem d.f - 2 and d.f.r:27 .
,
Locating at these d.f s, one finds that the required Fratio at the 0.05 level is 3.35
and at the 0.01 level is 5.49. Since the obtained value of F ratio is 8.29, which
exceeds 5.49, we reject the null hlpothesis and conclude that there is an overall
difference between the three groups of subjects on the educational achievement
test.

Ibsts after the F test

A general limitation of the F test
is that it only tells about the overall difference
between the groups under study but tells nothing about the location ofthe exact
difference. For example, in the above problem, the obtained Fratio of 8.29 is
significant, which definitely indicates that there is significant differerrce between the
groups under study, but whether the significant difference is betweenl and, B or A
and C or,B and C, cannot be said. Therefore, when F ratio is significant, one
needs some additional tests aft er the F test. Winer (197 l) has summarized half a
dozen different statistical tests, which can be used as additional test . One convenient
such statistical test is to apply the t test between the possible pairs ofthe groups.
Scheffe (1957)has introducedone suchrigoroustest, whichreducestheprobability
ofmaking a Type I error. Following the Scheffe technique, one can locate the
difference between tlrree means. Since there are three groups, thnee comparisons
are likelyto be made, namely,Avs. B,Avs. CandB vs. C. Followingthe Scheffe
technique, one is required to compute the Fratio with the help ofequation given
for each of three groups.

(M,- Mr)'
F_ sDi(Nt+ Nr)l N,N, ...(2.e)

Now, three{ fortlreepairs ofdistributioncanbe calculatedas follows: Fratio

for distributions I and .B:

(s6.10- ts.40)2
F_ (10 * lo)
= 16.31
soz.rgs
(10x10)

Self-Instructional Material 45
Meaning and Definition
a Research Problem
of F ratio for distributions I and C:

(s6.lo -31.20)2
= 6.10
s07.88sge!o)
NOTES (10x10)
F ratio for distributions .B and C:

(15.40-3 r.2o)2
=2.46
5s7.3359q*--1-E
(10x10)

As seen out earlier, F atthe 0.05 level ofsignificance for d.f.l :2 and d.f.
2:27 is 3.35. This value, ifmultipliedbyK- 1, yields (3 - 1) (3.35):6.70. Only
theFratio fordistributions,4 andB isgreaterthan6.T0. Hence, it isconcluded
that there is a sigfficant difference between the means ofAand.B only. The mean
difference between.,4 and C and B is not significant.

(iii) Pearson r
Ofall the meixures ofcorrelation the Pearson r, developed by Prof Karl Pearson,
is one of the most common methods of assessing the association between two
variables under study. It is also known as Pearsonproduct-moment correlation
and abbreviated to r. The size ofPearson r varies from *1 through 0 to -1 . In frct,
allcorrelationcoefficientshavethelimitof+l and-l.Acoefficientof+l indicates
a perfect positive correlatioq and a coefficient of-l indicates a perfect negative
correlation. The coefficient ofcorrelation indicates two things. First, it indicates
the magnitude ofa relationship. Acorrelation coefficient of say, +0.90 or-{.90
gives the same information about the magnitude or size ofcorrelation. The sign
makes no variation in the size ofthe correlation. Second, it gives an indication
regarding the directionofthe correlationcoefficient. Apositive correlationindicates
a similar trend ofrelationship between two variables, that is, as one increases the
other also increases or as one decreases, the other also decreases. Consider the
relationship between intelligence test scores and classroom achievement. Generally,
as intelligence test scores become higher, classroomachievement also becomes
better. Therefore, the direction ofthe correlation between these two variables is '
positive. Similarly, as fatigue increases, output decreases. Here the relationship is
negativebecause asone increases, the otherdecreases. Sometimes, therelationship
is not consistent. And in this situatior! coefficient of correlation is likely to be zero.

The Pearson product-moment correlation has two important assumptions:

(i) The relationship between X and )z variables should be linear. A linear
relationship refers to the tendency ofthe data, when plotted, to follow a
straight line as closely as possible. Although there are some statistical
treatnrents which can test whether or not the relationshfu is linear, generally this
is determined by the inspection ofthe scatter diagram or correlation table.

46 SelfJnstructional Material
 Meaning and Definition of

wWW
a Research Problem

NOTES
(a) (b) (c)

Fig, 2.3 (a) and (b) Non-homoscedastic and (c) Homoscedastic and Linear

(if The second assumption is ofhomoscedasticity (homo means 'like' and

scedasticity means scatteredness). Defined statistically, one can say that
homoscedasticityrefers to the fact that standard deviations (or variances)
for columns and rows in the scatter diagram are equal or nearly equal.
Figure above ilhrstrates homoscedastic and non-homoscedasic distnbutions.
In this figure, we have three diagrams. In Fig. 2.3 (a), the variance ofthe
distribution near the centre is smaller than the variances near both extrernes,
and hence, the distribution is non-homoscedastic. In Fig. 2.3 (b), the
variances near the bottom extreme are lower than the variances at middle
or at the top extreme, and , therefore, the distribution is no-homoscedastic.
InFig. 2.3 (c) the variancesare equalthroughout aswellas linear.
Pearson r canbe calculated by several formulas. One can calculate the
Pearson rbythe raw score formula or machine formula. The equation is:

N2 XY _>, XLY
...(2.10)
N>x2 -G&'][N>Y2 -Gn']
where r: Pearsonproduct-moment correlation coefficient; N:number of
scoresX: scores inXvariance; and I:
scores in lvariable.
The table below presents the scores of 10 students who were administered
intelligence test (X) and anxiety test ( Y). Pearson r has also been calculated from
the ten pairs of scores. The significance ofthe obtained r is tested with the help of
a table (Downie & Health, 1970:378). The obtained value of r as less than the
value required at even 0.05 level of significance. Hence, the null hypothesis is
accepted and it is concluded that the scores on intelligence test andthe scores on
the anxiety test are not correlated. The sign makes no difference in the magnitude
ofthe correlation.
When the data are arranged in bivariate distribution as is the case in the
scatter diagram or in the correlation table, Pearson r should be calculated by the
followingformula:

Ex'y'
= C*C,
N ...(2.11)
(o,,)(o,,)

Self-lnstruclional Material 47
Meaning and Definition of where r : Pearson r x' : deviation of scores from mean on x test;/' : deviation
a Research Problem
of scores from mean on / :
tes| /y' sum of frequencies ; C, : correction in x-series
scores; *d Cr: coffection iny-series scores.
Pearson r by raw-score method
NOTES
x Y x2 Y2 XY
10 8 100 @ 80
5 20 25 400 100
6 l5 36 22s 90
J l3 9 169 39
8 t6 g 256 1;8
t2 20 144 400 240
l3 t3 169 169 169
20 1l 400 tzt 120
l5 l0 225 100 150
l0 t2 100 144 120
>x:102 Er: 138 2X:1272 2Y:2048 ,,xY:1336
I,{2 XY -Z X >Y

(10x1336) - (102x138)
: .,/11r o;1r z i z) - (102)'z1 t(l 0X2048, - (r-a)1
-716 -716
J3325776 t823.671
df:N-2:10-2:8

CrmcxYouR PRocRESS

10. When is the students 'l' test orztest used?

11. What is the meant by degrees of freedom?
12. What is experimentalvariance?
13. What is error variance?
14. DefinethePearsonr.

2.6 TESTING HYPOTHESIS II: NON.PARAMETRIC

TESTING

The important non-parametric statistics are as follows:

(i) Chi-square(f)Test
(if Mann-WhitneyUtest
48 SelfJnsntctional Moterial
 (O Rank-difference methods (both fl and T) Meaning and Definition of
a Research Problem
(iv) Coefficient ofconc ordance (ll)
(v) MedianTest
(vi) Ikuskal-Wallis ^I{Test NOTES
(vi| Friedmantest

(i) Chi-square Qf) test

The chi-square is one ofthe most important non-parametric statistics, which is
used for severalpurposes. For this reasorq Guilford (1956) has called it the general-
purpose statistic. It is a non-parametric statistic because it involves no assumption
regarding the normalcy ofdistribution or homogeneity ofthe variances. The chi-
square test is used when the data are expressed in terms of frequencies of
proportions orpercentages. The chi-square applies onlyto discrete data- Howeveq
any continuous data can be reduced to the categories in such a way that they can
be treated as discrete data and then, the application ofchi-square is possible. The
formula for calculating .P is given below:
(f,--f")'
-2 - Y
LL.f" ...(2.12)

where : chi-square;f,: obtained or observed frequency; andf": expected

f
frequency or theoretical frequency. There are several uses ofthe chi-square test.
First, chi-square maybe used as a test ofequalprobabilityhypothesis. By
equalprobabilityhlpothesis one means theprobabrlityofhaving thefiequencies in
all the given categories as equal. suppose, for example, 100 students answer an
item on an attitude scale. The item has five categories of response options-
strongly agree, agree, neutral, disagree and strongly disagree. According to the
equal probability hypothesis, the expected frequency of responses given by I 00
students would be 20 in each. The chi-square test would test whether or not the
equal probability assumption is tenable. If the value of the chi-square test is
signfficant, the equal probability hypothesis becomes untenable and ifthe value of
the chi-square is not significant, the equal probability hlpothesis becomes tenable.
The second use ofthe chi-square test is in testing the significance ofthe
independence hypothesis. Byindependence hypothesis is meant that one variable
is not affected by, or related to, another variable and hence, these two variables
are independent. The chi-square is not a measure ofthe degree ofrelationship in
these conditions. It merelyprovides an estimate ofsome frctors other than chance
(or sanpfu error), which account for the possible rehtionshf. Generally, in dealing
withdatarelatedto independent hllpothesis,theyare first arrangedinacontingency
table. when observations on two variables are classified in a two-way table, data
are called the contingency data and the table is known as the contingency table.
Independence inacontingerrcytable existsonlywheneachtallyexhibits a different
event orindividual.

Self-Instructional Material 49
Meaning and Definition of The third important use of chi-square is in testing a hlpothesis regarding the
a Research Problem
nomal shape ofa frequency distribution. When chi-square is used in this connectior!
it is commonly referred to as a test of goodness-otf,t.
The fourthuse ofchi-square is in testing the significance of several statistics.
NOTES
For example, for testing the significance of the phi-coefficient, coefficient of
concordance, and coefficient ofcontingency, one converts the respective values
into chi-square values.
Ifthe chi-square value appears to be a significant one, one should also take
theiroriginalvaluesas significant. To illustratethis, a3x3 contingencytable, which
shows data of200 students who were classified into three classes onthe basis of
their educational qualification is used. The students' educational attainments are
measured in the course oftheir studyby classiffing them as superior, average or
inferior.

The use of Chi-square in a 3 x3 Contingency Table

Superior Average Inferior

Master 30 (25) ls (1s) s (10) 50
Bachelor 2s (2s) l0 (rs) ls (10) 50
Intermediate 4s (s0) 35 (30) 20 (20) 100

100 60 40 200

(The figures inbrackets show expected frequency)

The question posed is: Is educational achievement related to educational
qualification? The obtained data have been shown above. The first step in
calc:ulmltngP as a test of significance of independence or the relationship between
educational achievement and educational qualification is to cornpute the expected
frequency. The nullhypothesis is that these two variables are not related or are
independent, and ifthis hypothesis is true, the expected frequencies should be as
follows:

Cells of Table Expected Frequency

Upper left (100x50) 1200=25
Upper middle (60xs0) 1200:ts
Upper right (40x50) 1200: l0
Middle left (100x50) 1200:2s
Middlemiddle (60x50) 1200:15
Middleright (40x50) 1200:10
Lower left (100x100) 1200:50
Lower middle (60x100) 1200:30
Lower right (40x100) 1200:20
2=200

50 Self-Instructional Material
Aftercalculatingexpected frequencyforeachce[ the chi-squarernaybe calculated Meaning and Definition of
a Research Problem
as:

(f" - f")'
.f" .f" f,- f" (f, - f)' f" NOTES
30 25 +5 25 I
l5 l5 0 0 0
5 l0 -5 25 2.5
25 25 0 0 0
10 l5 -5 25 1.67
l5 10 +5 25 2.5
45 50 -5 25 0.5
35 30 +5 25 0.83
20 20 0 0 0
2f,:2oo z,f":2oo E(f"- "f") :o E:9.00
df: (r- lX,K: 1) : (3 - 1X3 - : 1) 2x2 : 4

Entering the probability table of chi-square, one finds that the value ofchi-
square for d.f :4 atthe 0.05 level should be 9.488. As the obtained chi-square
is below it (p >> 0.05), one concludes that the null
fulpothesis is retained. Hence,
thetwo variables, namely, educationalqualificationandeducationalattainment in
the present studyare found to be independent. For calculating d.f in a chi-square
test, the formula as noted above is (, - r) (k - I ) where r : the number of rows
and fr is the number of columns.

chi-square: when the data has been arranged in a 2x2 contngency table
(where d.f. : 1), we need not calculate the expected frequency in the manner
described above. In such a situation, the chi-square can be directly calculated
with the help ofthe following equation:

Nil AD - BC ll'
x2: (A+ B)(C + D)(A+C)(B + D) ...(2.13)

where A, B, C and D are symbols for frequency of four cells in a zx2 table; N
: total number of frequencies; bars (II) indicate that in subtracting BC fromAD
the signis ignored.

Suppose the researcher wants to know whether or not the two given items
in the test are independent. Both items have been answered in 'yes'or 'No'forrn
The test was administered to a sample of400 students and the obtained data were
asfollows:

Self-Instructional Material 51
Meaning and Definition of Chi-square in a 2x2 Table
a Research Problem
Item No. 6

No

NOTES No 180 120 300

Item No. l0 A B

Yes
90 l0 100

C D

270 130
According to the formula:
d.f : (r - l)(k - I) : (2 - 1)(2- 1) : 1

Entering the probability table ofchi-square, we find that for d.f , the value
of chi-square at the 0.00 I level should be 10.827 . As the obtained value of the
chi-square is much above it, we conclude that item nos. 6 and 10 are not
independent, that is, theyare related.
Sometimes it happens that with I d.f., any one of the expected cell
frequencies becomes less than 5. In such a situation a correction called Yates'
correction for continuity is applied. Some writers have suggested that Yates'
correction for continuity should be appliedwhen anyofthe expected frequencies
goes hlow 1 0. Where frequencies are large, this correction makes no difference
but where frequencies are small, Yates'correction is significant. Yates'correction
consists in reducing the absolute value ofdifference betweenf,ardf"by 0.5, that
is, each fo which is larger than fe is decreased by 0.5 and each fo which is smaller
than fe is increased by 0.5. The formula for chi-square in such a situation is as
givenbelow:

xlr eo- BCt-+1'

t: (A+ B)(C + D)(A+C)(B + D)
...(2.r4)

where subscripts are defined as usual. Suppose, 60 students (50 boys and 10
girls) were administered an attitude scale. The items were to be answered in 'Yes'
and'No'forrn
Their frequencies towards 10 items are presented in the table below. The
question is: Do the opinions ofboys and girls differ significantty?

Chi-square with Yqtes'Correction in a 2x2 Table

Yes No

Boys 2A 30 50

A B
1 10
3
Girls
C D
23 JI 60
52 Self-Instructional Material
According to Equation (2.14): Meaning and Definition of
a Research Problem

t:60 (sOxl0x23)(37) NOTES

_ 60l140-90l-301'z _ 60x400 24000

= 0.056
425500 42s500 425500
In the above example, the expected frequency (23x10160) is less than 5.
Hence, chi-square has been calculated byEquation (2.14). Entering the table for
chi-square, we find that for d.t. : l, the value of chi-square at the 0.05 level
shouldbe 3.841. since the obtainedvalue is less thanit (F> 0.05), one concludes
that the opinions ofboys and girls do not differ significantly.

(ii) Mann-Whitney U test

The Mann-Whitney Utest is a non-parametric substitute for the parametric /test.
This test was independently proposed by Mann and whitney. The Mann-whitney
Utest is used when the researcher is interested in testing the significance ofdifference
between two independently drawn samples or groups. For applications ofthe U
test it is essential that the data be obtained on ordinal measurement, that is, they
must have been obtained in terms ofrank. Where the data have been obtained in
terms of scores, for application ofthe Mann-Whitney [./test, it is essentialthat
those scores be converted into rank without much loss of information. It is not
necessary for the application of the Mann-Whitney U test that both groups must
have unequal size. However, this test canalso be applied to groups having equal
size.

Here the calculation the Marur-whitney utest, which are concemed with
larger sample size ofmore than 20 cases is given:

rJ: N,N,*{94-r^, ...(2.1s)

(J: N4z,'e+-!-x^, ...(2.16)

Table below presents the scores oftwo groups on the Lie scale. Group I
has 10 subjects and Group II has 2l subjects. The first step is to rank all the
scores in one combined distribution in an increasing order of size. The lowest
score (taking both sets ofscores together) is 7 (second column) and hence, it is
given a rank of 1. The nest score is 8, which is again in the second column and it
has been given a rank of 2. The third score from below is 1 0 (in the first column),
which has been given a rank of 3. In this way ranking is continued until all scores
receive ranks. Subsequently, the two columns ofranks are summed. At this point,
a check on arithmetical calculation is irrposed. The check is that the sums ofthese
two columns must be equal to N (N + 1)12.
Self-lnstuctional Material 53
Meaning and Definition of (lOW+t) (31X32)
a Research Problem check: x& +tR2 = = 88.5 +407.5 = 496; = 496
22
Hence, we can proceed:

NOTES (byEquation2.ll)

(J : N,N,.{ga!-xR, : (t0x2l).q!rya!-83.s =176.5

(byEquation2.I6)
Nz({u +l)
(J : NtNz* : (l0X211+9!Q1{)
2"2 -xR2 -407.s=33.5

Table 2.3 Calculotion of the Mann-llhitney U Tbstfrom Larger Sample Sizes

Gr.I Gr. II Rr R2
(N,:10) (N,:21)
l8 32 7 13

t4 40 15 l8
30 3l ll t2
a
10 39 J 16.5
39 15 16.5 6
26 8 9 2
27 47 10 t9
19 33 8 t4
35 52 l5 22
1l 48 4 20
7 I
50 2t
6t 27
58 24
53 23
59 25
60 26
65 30
63 28
67 3l
& 29

XR, : 88.5 tR":407.5

It of Utest that one wants. For testing the significance of

is the lower value
the obtained U, its values are converted into z score as shownbelow:

54 Self-lnstructional Material
 Meaning and Definition of
u -N,N, a Research Problem

m
!12
...(2.17)

NOTES
176.s _(1ox2l)
2 71.5
=3.02
(l0X2lxl0 + 21+ 1) 23.664

A z score from +1 .96 to +2.58 is taken to be significant at the 0.05 level of

significance and ifthe z score is greater than even +2.58, one takes it to be significant
at the 0.01 level. Since the obtained z is3.02, on can take the value ofthe Mann-
Whitney Uto be a significant one. B-ejecting the null hypothesis, one can conclude
that the trvo groups differ significantly on the measures ofthe Lie scale.

(iii) Rank-difference methods

The methods of correlation based upon rank differences are very common among
behavioural scientists. These are two most common methods which are based
upon the differences in ranks assigned on the X md l/variables. One is the Spearman
rank-difference method andthe other is the Kendallrank-difference method. The
Spearman rank-difference method symbolized by fl (read as rho) is a very popular
method ofcomputing the correlation coefficient between two sets ofranks or
between two sets of scores converted into ranks. The method has been named
after Spearmanwho discovered it. This method is applicable whenthe number of
pairs ofscores or ranks is preferably small, that is, 30 or below.
The equationis:

oFp2
p:,_r(N_,
1L/
...(2.18)

where p: Spearman's rank-difference correlation coefficient; D: difference

between rankl andrar*2; and N: Number ofpairs ofranks or scores.
To illustrate the calculation offf consider the data given in table that follows,
which show the scores of 12 students on the intelligence test (.f,) as well as on the
educational test (Y). The first step is to rank both sets of scores separatelygiving
the highest score a rank of 1, the next highest score a rank of2, and so on. Then,
keeping the algebraic signs in view, the difference between two sets ofranks is
conputed. ThisisnotedundercolunmD. Subsequently, eachdifferenceis squared
and noted under columnQ. Substituting the values in the equation, we get aft of
-0.185. Following the table ofGuilford (1956: Table L, p.549), we can test the
significance ofthe obtained fr. Since the obtained value of fi is less than the value
given at the 0.05 level (fi > 0.05) for N= 12, one can accept the null hypothesis
and can conclude that Xan d Y are independent and whatever correlation has been
found is due to the chance factor.

Sel/-lnstructional Material 55
Meaning and DeJinition of Table 2.4 lllustrqtion of the Speorman Rank-Diflbrence Correlation
a Research Problem

x Y Rank, Rank, D (fi, - R,) D2

47 68 8.5 I +7.5 56.25
NOTES 50 60 5.5 2.5 +3 9.00
70 54 2 7 -5 25.00
72 53 1 8 -7 49.00
46 CI 10 2.5 -7.5 56.25
50 55 5.5 6 -0.5 0.25
42 48 ll 9 +2 4.00
58 30 J t2 -9 81.00
55 45 4 l0 -6 36.00
36 43 12 11 +l 1.00
49 59 7 4 +3 9.00
47 56 8.5 5 +3.5 12.25
XD: 0.0 2D2 = 339.00

6(339)
- , 2034
'p:l_ t2(122-t)
=l__=l_1.195=_0.195
t7 t6

Another method of computing the rank-difference correlation has been

developed by Kendall. The method is known as Kendall's r for which the formula
is as follows.

r : ...(2.1e)
0r2)N(N )
where r: Kendall's r, ^S: actual total; and N= nurnber of objects or scores
which have been ranked.
Suppose 12 students have been administered two tests and their scores are
presented inthe following table.

Scores of l2 Students on X and Y kst

A B C D E F G H I J K L
X 20 26 t7 l6 l5 23 22 24 19 28 30 l0
Y 70 80 40 45 38 49 77 76 72 47 36 35

The first step is to rank both sets of scores giving the highest score a rank of
1, the next higher a rank of 2, and so on. The following table presents the ranks
based upon two sets ofthe scores given in that table above. Subsequently, the
ranks ofthe Xtest are reuuranged in a way that they appear in a natural order like
1,2,3.
Ranl{s based upon Two Sets of Scores given in the aboye Table

A B C D E F G H I J K L
X 7 J 9 l0 ll 5 6 4 8 2 I t2
1
Y 5 I 9 8 l0 6 2 J 4 11 12

56 Self-lnstructional Material
 Accordingly, ranks onthe )ztest are adjusted. The followingtable presents Meaning and DeJinition of
a Research Probtem
the ranks in a rearranged order. Subsequently, the value of,S is computed. For
this, we start with the rank on the Iztest from the left side.

Rearranged Order of Ranks

NOTES
A B C D E F G H I J K L
x I 2 3 4 5 6 7 8 9 l0 ll t2
Y lr 7 I 3 6 2 5 4 9 8 l0 12i
The first rank on the left side is 11. count the number ofranks which are
above I 1 and the number ofranks which are below I l, separately. only one rank
(that is, l2) falling at the right ofthe first rank on the rtest is above I I and the
remaining 10 ranks fallbelow it. Hence, its contribution to,Swouldbe equalto
I - 10. Likewise, the second rank on the Itest is 7. The four ranks falling right of
7 , are above 7 and,6 ranks are below it. Hence, its contribution to s would be

4-6.
Identical procedures are repeated for other ranks on the lztest. Thus:
s=(1 - 10)+ (4-6)+(7 - l)+ (4-3 )+(6-0)+(4_ 1)+(4_0)+
(2 - r) + (2 -0) + ( I - 0) : (-e) + (-2) + (e) + (6) + ( 1) + (6) + (3) + (4) + ( I
)
+ (2) + (1) = 3l * 1l :22.
Substituting this in the formula given in Equation (2. 1 9) :

s22))
' =- =-=n
.^lN(N-r) lnoz-rl 66
222

22
, the formula for
follows:
]

whichisas I

...(2.20\
2(2N +s)
9N(,^/ - 1)

Hence

0.33 0'3
ffi=m
0.33
z: =0.2209 = 1.4938
e(12)(12-t)

Since the obtained z score is less than 1.96, one can say that this is not
significant even at the 0.05 level. Accepting the null hypothesis, one can say that
the given set of scores is not correlated. According to Siegel ( I 956), r has one
advantage over p, and that is that the former can be generalized to partial correlation
Ifboth r and p are computed from the same data, the answer will not be the same
and hence, nulirerically, they are not equal.

Self-lnstructionolMaterial 57
Meaning and Definition of (iv) Coeffi cient concordance, ll/
a Research Problem
The coefficient ofconcordance symbolized by the letter Whasbeen developed by
Kendall and is a measure of correlation between more than two sets of ranks.
Thus, W is ameasure of correlation between more than two sets ofrankings of
NOTES
events, objects and individuals. When the investigator is interested in knowing the
inter-test reliabiliry lZis chosen as the most appropriate statistic. One characteristic
of l/which distinguishesit fromothermethods ofcorrelationisthat it is eitherzero
or positive. It cannot be negative. W canlcr- computed with the help ofthe formula
givenbelow:

W_ ...(2.2r)
arr(,a,r, - lr)
t2
where I/: coefficient ofconcordance; S: sumofsquares ofdeviations fromthe
mean ofR ; K: number ofjudges or sets ofrankings; and N: number ofobjects
or individuals which have been ranked. Suppose four teachers (A, B, C and D)
ranked 8 students on the basis ofperformance shown in the classroom. The ranks
given by the four teachers are presented in the following table. The details of the
calculations have also been shown.

Ranks given by Four Tbachers to Eight Students on the

bosis of Classroom Performance

Students
Teachers
(i) (ii) (lIl) (iv) (v) (vi) (vii) (viii)
A 3 4 7 5 8 6 '2 I
B 2 3 6 4 8 7 I 5

C I 3 5 6 8 7 2 4
D J 4 2 5 7 6 I 8

RJ 9 t4 20 20 31 26 6 18

9 +14+20+20 +31+26+6+18 144

MeanofR.l =18
88
s : (9- l8)2+ (14 - lS)'z+ (120- 18)'+ (20- l8)2+ (31 - 18)2
+ (26 - 18)2 + (6 - 18)' + (18 - 18)'
= (- 9)2 + (- 4)' + (2)'+ (2)' + (13)'? + (8)' + (-12)2 + (0)' :482
Now substituting in Equation (2.2 I ):
482
W_ -482 =0.717 =0.72
672
fit+t'ta'-al
when N> 7, the signfficance of Zis tested by converting its value lrirto N with the
help ofthe following equation: *:
K (N- l) WThns, X:4
(8 1) x 0.72 -
:20.16 and d.f in this situation is always equalto N- 1. Hence d.f. 8 I:7 . : -
5E SelfJnstructional Material
 From the probabilitytable for chi-square we find that the value of chi-square for Meaning and Definition of
d.f. :7 at 0.05 level of significance should be I 8.475. Since the obtained value of a Resenrch Prohlem

the clri-square exceeds this required value, one can take this value of W as a
significant one. Thus rejecting the null hypothesis, one can say that there is an
overallsignificant relationship in ranking done bythe fourteachers. NOTES
(v) Median test
The median test is used to see iftwo groups (not necessarily of same size) come
from the same population or frompopulations having the same median. In the
median test, the null hypothesis is that there is no difference between the two sets
of scores because they have been taken from the same population. If the null
hypothesis true, halfofthe scores in both the groups should lie above the median
is
and the rernaining half ofthe scores should lie below the median. The following
table presents the scores oftrvo groups of students in an arithmetic test. The firsi
step in the comptttation of a rnedian test is to compute a common median for both
distributions taken together.

Scorcs of30 Sndents on an Arithmetic Tbst

Gr. B
16, 17, g, 12, 14, 9,7, 5,20,22, 4,26,27,5, 10, lg
(1/: l6)
Gr. B
(1/= 14)
28, 30, 33. 40, 45,47,40, 3g" 42, 50,20, lg,lg, lg

For computing the comnron median, both the distributions are pulled together
as shown in the following table:

Scores

49*53 I

1418 2

3913 -)

34- 38 I

29*23 2 Median -ft(Nlz-F)i

24-28 .f^
3

t9-23
(30 t 2- l3)s
5 = 18.5 +
l4- l8 5

9-13 =20.5
-1

4-8 5

N: 30

Subsequently, a 2x2 contingency table is built as follows:

Now, the chi-square test can be applied. For computing chi-square from
a2x2 table, we may follow the equation for chi-square (12). yates, correction
is not needed here because none of the cells contain an expected frequency less
than 5.
Self-lnslructional Material 59
Meaning and Definition of Above Not atrove
a Research Problem Mdn Mdn

A B l6
Gr. A
J l3
NOTES 14
C] D
l0 4

t3 t7 30

Now substituting the values in Equation (2.13) we get:

-
x':@ 3Oil 3(4) - ( l3Xl o) ll:

30[ l2 - 130 l]'? 4t7720

= 8.438 =8.44
49504 49504
df : (r- lXc- 1) : (2 - 1\(2- l) : 1

Fromthe probabilitytable for chi-square, we find that for d.f 1 the chi-
:
square value at the 0.01 level should be 6.635. Since the obtained value ofthe
chi-square exceeds this value (p < 0.01), we canreject the nullhypothesis and
conclude that the two samples have not been drawn fromthe same population or
from populations having equal medians.

(vi) Kruskal-Wallis 11 test

The primary difference between the F test and the Kruskal-Wallis Fltest on the
one hand and the Friedman test on the other hand is that the Ftest is a parametric
analysis ofvariance, whereas the Htest and Friedman test are the non-parametric
analysis. The .Fl test is a one-way non-parametric analysis of variairce and the
Friedman test is a two-way non-parametric analysis ofvariance.

Tabte 2,5 It Test./'ront Scores Obtainecl by Tht"ee Groups on Lie Scale

Gr.A (N= 6) Gr. B (N= 8) Gr. C (N= 10)

l5 (14) t7 (ls) 20 (17)
10 (e) 9 (8) 25 (le)
8 (3) 8 (6) l3 (12\
5 (4) t4 ( l3) ll (10)
6 2) 2 (1) 26 (20)
4 8 (6) 24 ( 18)

t2 (l l) 36 (24)
t8 (16) 30 (23)
29 (22)
27 (21)
R;: 38 Ri:76 fti: 186

60 Sef-Instruclional Material
 The llltest is used when the investigator is interested in knowing whether or Meaning and Definition of
a Research Problem
not groups ofthe independent samples have been drawn fromthe same population.
Ifthe obtained data does not fulfill the two basic parametric assumptions, namely,
the assumptionsofnormalityand the assumption ofhomogeneity ofvariances, the
.F/test is the most appropriate statistic. The equation for the lltest is as given NOTES
below:

H: ,(fu[r#]-3(N+,) ...(2.22)

Where N: number in all samples combined; R: sum ofranks in7 sample; *d N,

: number in7 sample.
Data to illustrate the calculation of the /1test have been given in above
table. Three groups of students were administered a Lie Scale and their scores are
presented in table above. The first step is to combine all the scores from all ofthe
groups and rank them with the lowest score receiving a rank of I and the largest
score by rank N. Ties are treated in usual fashion in ranking subsequently, the sum
ofranks in each group or column is found and the l1test deternrilres where these
sums ofranks are so disparate that the thlee groups cannot be regarded as being
drawn from the same population. The ranks assigned to each score eamed by the
member ofthe group are given in brackets. Now, substituting the value in the
equation given above, we get

p: _LL_.[:qq .06)'.qyll _3e4+t)

(24)(2s)l- 6 8 l0l
: 53067
ooo
'192
-75 =88.445 -75 =13.445 =13.44

When each sample has six or more than six cases, the lltest is interpreted
as chi-square. In such a situation, d.f. : number ofgroups or samples minus one.
So, here d.f. :2. Entering the probability table for chi-square, we find hat for d.f
:2,the value ofchi-square at the level ofsignificance should be 9.210. Since the
obtained value oflltest exceeds this required value, it can be said that the F/value
is significant. Rejecting the null hypothesis, one concludes that the samples are
independent and that they have not been drawn from same population.

(vii) Friedman test

As mentioned earlier, the Friedman test is a two-rvay non-parametric analysis of
variance. When the groups are matched doubts exist about the trvo basic parametric
assumptions, namely, the assumption of norrnality and the assumption of
homogeneityofvariances, the investigator resorts to the Friedman test for testing
whetherornotthe sanpleshavebeendrawnfromthe samepopulation. To illustrate
this, the calculations ofthe Friedman test have been presented in table below. The
following equationcalculates the Friedrnantest is as follows:
Sclf-lnstructional Malerial 6l
Meoning and DeJinition of
a Research h'oblem
xz :
A, -:_(R,),
NrK(K + l) '
-3//(K + r) ...(2.23)

where X2: Friedman test;y'y': number ofrows; K: number ofcolumns; and R-

NOl'ES : separate sums ofranks ofeach colunm.

Table 2.6 Friednnn Tbst of Tiyo-,tta.v ANOVA /rom Scorcs

of Thrce Groups on Recall Tbst
I u III Iv v
Gr. A r2(4) E(2) 4(l) l0(3) r 6(s)
Gr. B 10(3) 7(2) 6(l) r l(4) r 7(5)
Gr. C 7(2) 83) 4(l) l 2(s) r 0(4)
Rj 9 7 -) t2 l4

The fir'st step in calculation of the Friedman test is to rank each score in
each row separately giving the lowest score in each row a rank of 1 and the next
lowest score in each row a rank of 2, and so on. The ranking can also be done in
reversed order, that is, giving the highest score in each row a rank of l, the next
highest score in each row a rank of2, and so on. Rank assigned to each score in
each row i;s given in parentheses. The Friedman test is applied to determine whether
or not the rank totals symbolized by R- differs significantly. Now, substituting the
values inthe equation ofchi-square, we get

v2
.t?
-
-' I\ +(14):]-3(3X5+l)
fror: -/7\r -/3):
^r- (3X5X5+l;Lvr '\"
: 63.866 - 54:9.87
When the number or rows (Af and the nunrber of columns (r(') are too
small, the significance of the Friedman test can be ascertained with the help of
special tables (Siegel, 1956). For example, when K:4, N:2 to 4 or when
K:3,N:2to 9, the significance ofthe Friedrnan test can be done through these
special tables. But when the number of rows and the number of columns are
greater than those said above, the Friedman test is interpreted as the chi-square
test. In the present example, the significance of the Friedman test would be
interpreted in terms ofchi-square. The d.f, is always equal to 1(- 1 for chi-square
applied as a test of significance of the Friedrnan test. Hence d.f. in the present
example would be K - I = 5 * | : 4. Entering the table for d.f 4, we find that the
chi-square should be 9.488 at the 0.05 level ofsignificance. Since the obtained
value of the Friedman test exceeds this value (p < 0.05), one rejects the null
hypothesis andconclude that th,ree matched groups differ significantly.

2.7 CORRELATION AND REGRESSION

Correlational methods are the most commonly uscd statistical techniques in the
testing field. Some important methods of correlation are: Pearson r, Spearman
rank difference method and Kendall's Tndiscussed earlier. Acorrelation coefficient
is a mathematical index that describes the direction and magnitude ofa relationship.
62 Se(-lnstructional Marerial
There is also a related technique called regression which is used to make a prediction Meaning and Definition of
a Research Problem
about scores on one variable on the basis ofknown score on another variable. In
fact, these predictions are done from the regression line, which is defined as the
best-fitting straight line through a set ofpoints in a scatter diagram. It is estimated
by using the principle of least squares which, in fact, minimizes the squared NOTES
deviations around the regression line.
This canbe explained through an example. The mean is the point ofleast
squares for any single variable. In other words, the sum of squared deviations
around the mean will be less than it is around any value other than mean. For
example, the mean of the five scores namely, 2,3, 4,5 and 6 is ...X/N = 2015
:4. The squared deviation of each score around the mean can now be easily
determined. For score 6, the squared deviation is (6 - 4) x 2: 4; for score 5, the
squared deviation is (5 - 4) 2 :1. The score 4 is equal to the mean and
"
therefore, the squared deviation around mean will be (a - a) x 2:0. Thus, by
definition, the meanwill always be the point least squares. The regression line is
the line ofleast squares or ruining mean in two dimensions or in the space created
bytwo variables.
As descnbed earlier, a regression line is the best-fitting straight line through
a set of several points in a scatter dragranl which is the picture ofthe relationship
between two variables. The regression line is described as a mathematical index
called regression equation. The general linear regression equation for the straight
lineis:
Y: bx + a (figure 2.4) ...(2.24)
Where a is the intercept, that is, value of Iwhe n X rs zero.In other words,
the point at which the regression line crosses the )z-axis (a) is found by using the
followingformula:

34 67
Fig.2.4: Y = b2+a
a:Y-bx ...(2.2s)
Self-Inslructional Material 63
Meaning and Definition of D is the slope of regression line or it is called the regression coefficient. It is
a Research Problem
expressed as the ratio of sum of squares for the covariance to the sum of
squares for In the above figure, two regression lines have been drawn based
-X
on equation Y: a + Dx. The regression equation gives a predicted value for
NOTES I, on the basis ofX. This predicted value is called Y'.lnfact, the actual score
and predicted score on )'is rarely exactly the same. Some difference between
the observed and the predicted score, that is, Y- Y' occurs and this is technically
called the residual. Symbolically, the residual thus, is defined as Iz- I'. The
best-fitting line actually keeps the residual to minimurn [n other words, it minimizes
the deviation between the observed and the predicted Iscore. Since residuals
can be either positive or negative and will cancel to zero if averaged, the best-
fitting line is rightly estimated by squaring each residual. In this way, it can be
said that the best-fitting line is best obtained by keeping these squared residuals
as small as possible.

Difference between correlation and regression

The correlation coefficient has a reciprocal nature whereas regression lacks this
property. The conelation berween X and l/willalwals be the same as the correhtion
between Y andX. For example, ifthe correlation between test anxiety and test
performance is 0.64, the correlation between test performance and test anxiety is
also 0.64. One cannot claim similar thing about regression. In fact, regression is
used to transform one variable into estimated scores on the variable. In other
words, regression is oftenused to predict raw scores on lon the basis ofraw
scores onX. For eiample, tluough regression equation, the investigator might be
interested in predicting score on the intelligence test on the basis ofscore on, say,
the anxiety test. The coefficient denoting regression ofXon lis usually not the
same as the coefficient denotingthe regressionYonX. Since regressionusesthe
raw scores ofthe variables, the trait ofreciprocity does not hold true.
Regression analysis tends to show how changes in one variable are related
to changes in another variable. In psychological testing, regression is often used to
determine whether changes in the test scores are related to the changes in
perforrnance. For example, do persons who score higher on mechanical reasoning
test perform better as a mechanical engineer? In fact, regression analysis and
related correlational methods show the extent to which these variables are linearly
related. Besides, theyalso often create an equation that estimates scores on the
criterion (such as performance as mechanical engineer) on the basis of scores on
a predictor (such as scores on mechanical reasoning test).

Choosing appropriate statistical tests

It is customary to choose the appropriate statistical tests on the basis ofthe nature
ofthe obtained data.
Ifthe data fulfills the requirement of parametric assumptions, anyofthe
parametric tests which suit the purpose can be selected. On the other hand, any of
64 SefJnstntctional Muterial
the non-parametric statisticaltests that suits thepurpose canbe choseq ifthe data Meaning and Definilion of
a Research Problem
does not fulfillthe parametric requirements. Other factors to be kept in mind in
While selecting the appropriate statistical tests other factors like the number of
independent and dependent variables and the nature ofthe variables-whether
they are nominal, interval or ordinai-should be kept in mind. NOTES
When both the independent variable and the dependent variable are interval
measures and are more than l, multiple correlation is the most appropriate statistic.
on the other hand, when they are interval measures and their number is only one,
Person r may be used. As has been noted earlier, with ordinal and nominal measures
the non-parametric statistics are the common choice. Sometimes, researchers
transform the measures so that the appropriate statistical test mayh applied with
out loss ofmuch information. For example, ifscores oftwo groups on interval
measures are available but the data does not fulfill the requirement ofthe /test, the
researcher can transform the interval measures into ordinal measures and
subsequently apply the Mann-Whitney test.

CrucrYouR PRocRESS

15. What isthe chi-squaretest?

16. Define the Mann-Whitney Utest.
17 . Define the Spearmanrank-difference method.
18. What iscoefficientofconcordance W
19. When is the freedman test of two-way analysis ofvariance used?

2.8 SUMMARY
o while carrying out research, defining the problem is the most important
step. Once this is done, the next step is formulating a hypothesis, which is a
tentative solutionto the problenr, inthe view ofthe researcher. Hlpotheses
are oftwo tlpes-solvable and unsolvable.
o There are different types of hypotheses which include causal hlpothesis,
descriptive hypothesis, simple hypothe sis, co mplex hypothesis, research
hlpothesis, null hypothesis and statistical hypothesis.
r To test the hypotheses, the following kinds oftests are used: parametric
tests and non-parametric tests.
o In a parametric statistical test, certain conditions about the population
parameter from which the sample is drawn, are specified. Examples of
paramedic tests are z test, ttest andFtest.
o on the other hand, in a non-parametric test, no conditions are specified
about the parameters of the population from which the sample is drawn.
Self-Instructional Material 65
Meaning and Definilion of o The appropriate statistical tests are chosen on the basis ofthe nature ofthe
a Research Problem
obtained data- Ifthe data fulfills the requirement ofparametric assumptions,
any ofthe parametric tests which suit the purpose can be selected. On the
other hand, any ofthe non-parametric statistical tests that suit the pu{pose
NOTES- canbe choserq ifthe data does not fulfillthe parametric requirements.
. Other factors to be kept inmind while selecting the appropriate statistical
tests other factors like the number ofindependent and dependent variables
and the nature of the variables-whether they are nominal, interval or
ordinal--should be kept in mind.

2.9 KEY TERMS

o Hypothesis: A proposition that can be tested and suggests a possible
solution.
o Research hypothesis: A working hypothesis which is derived from
researcher's theory.
a Null hypothesis: Ahypothesis that negates difference
a Statistical hypothesis: Ahypothesis where lr the null hypothesis (H) and
the alternative are expressed in numerical terms.
Parametric statistical test: Atest in which certain conditions, about the
populationparameter fromwhich the sample is drawn, are specified.
Non-parametric statistical test: A test which does not specify any
conditions about the parameters ofthe population from which the sample is
drawn.

2.I0. ANSWERS TO 'CHECK YOUR PROGRESS'

l. Research is an inquiry for gaining knowledge. It is a systematic search for
information on a chosen topic. The definition ofa researchproblem has to
be carefully scrutinized for the following factors: (D There should to a person
group or institution to wham a give problem can be attributed; (ii) Two
sources ofactionmust be possible; (iii) There must be at least two possible
outcomes: and (iii) the course ofaction must have a bearing on the outcorre'.
2. While selecting aresearchprobler4 a researcher should follow the following
principles: (i) Avoid an often researched topic; (iii) Avoid controversial
issues; (iii) Keep costs, time and competencies in mind; (iv) Choose a
problemofgenuine interest and (v) Carryout apilot study, ifpossible .
3. A research problem should be stated in operational terms. The time-frame
should be set and the scope and limitations ofthe study should be clearly
specified.

6 Eef-Instntctional Material
 4. Researchproblems are oftwo types: solvable andunsolvable. Meaning and Definilion of
a Research Problem
5. A hypothesis is a suggestion about a possible solution. It is a patented
association between variables, a conjecture about a relationship, and a
testable proposition.
NOTES
6. A null hypothesis is a hlpothesis that negates the research hypothesis. It is a
no difference hypothesis.
7. Statistical hypothesis is also known as the altemate hypothesis. This
hypothesis states that the two groups that are being studied, would differ.
8. In a parametric statistical test, certain conditions are specified about the
population parameter from where the sample is drawn.
9. A non-parametric test does not specify any conditions about the parameters
of the population, from which the sample is drawn.
10. when significance of difference between two means is to be tested the
students I test or t ratio or the z test or z ratio, is used.
I l. Degrees offreedom is the number ofobservations that are independent of
each other and cannot be deduced from each other.
I2 . Experimental variance is the effect ofmanbulation ofthe independent variable
on the dependent variable.

1 3 . An error variance is the variability in measurements that occur as a result of

factors that are beyond the control ofthe experimenter. In psychological
research individualdifferences, motivations, attitudes etc are typical factors
that contribute to error variance.
14. The Pearson r is a common correlation method where the size ofthe r
varies from + I - 0 + - I this gives the degree of association.
15. The chi-square is a generalpurpose statistic. This test is usedwhen data are
presented in frequencies, proportions or percentages. It can onlybe used
when the data are discrete.
16. The Mann-whitney U test is a test of difference between two means
independently obtained fromtwo groups or samples. The scores have to
be interms ofranks.

17. The Spearman rank-difference method is symborzed as fr (read as rho). It

is a commonmethod for calculating coefficient ofcorrelationbetween sets
ofranks
I 8. coefficient of concordance zwas developed by Kendall and is a measure
ofcorrelation between more than two sets of ranks, events or individuals-
this is used when one wishes to studythe inter-test reliability.
I 9. The Freedman test oftwo-way analysis ofvariance is useful when rnatched
groups andtheirparametric conditions are in question.

Self-lnstructional Material 67
Meaning and Detinition of
a Research Probtem 2.ll QUESTIONS AND EXERCISES

Short-Answer Questions
NOTES
l.List anytwo specifications for a researchproblem'
2. List two considerations while selecting a research problent
3. Define parametric statistical test.
4. Define non-parametric statistical test
Long-Answer Questions
1. Explain a research problem and hypothesis in detail '
2. Explain in detail the three most important parametric statistics: (i) Student's
t testandz test, (ii) FRatio, and (iii) Pearsonr
3. Explain sample selectioq types and size and the steps in sampling in detail'

2.I2 FURTHER READING

Guilford, J.P. and B. Fruchter. 1978. Fundamental Statistics in Psychologt
and E duc atior. New York: McGraw-Hill.
Kerlinger, F.N. 1973. Foundations of Behavioral Researcft. NewYork: Holt
Reinhart and Winston, Inc.
Kothari, C.R. 1 995. Research Methodolo gt-Methods and Techniques. New
Delhi: Wishwa Prakashan.
Singh,A.K. 2008. Tests, Measurements and Research Methods in Behavioural
Sciences. New Delhi: Bharati Bhawan.
Winer, B.J. 1971 . Statistical Principles in Experimental Design Tokyo:
McGraw-Hill Kokakusha.

6t Sel!-Instructional Mareriul
 Sampling Design

UNIT 3 SAMPLING DESIGN

Strucfure
NOTES
3.0 Introduction
3.1 Unit Objectives
3.2 SamplingDesign
3.2.1 Steps in Sampling Designing
3.2.2 Principles for Selecting a Sampling hocedure
3.2,3 Systemic Bias and Sampling Enors
3.2.4 Criteriaof Good Sample Design
3.2.5 Types of Sampling Designs
3.2,6 Types ofSampling
3.3 StatisticalDistributions
3.4 Sunrnary
3.5 Key Terms
3.6 Answers to 'Check Your Progress'
3.7 Questions and Exercises
3.8 Further Reading

3.0 INTRODUCTION

In this unit, you will learn how to design the right sample for a chosen study. All
items in a field ofinquiry are thought to constitute a universe or a population. A
sample is any number ofpersons selected to represent a population according to
some rule orplan. Hence, a sample is a smaller representation ofthe population. A
measure based on a sample is known as a statistic.

To measure the characteristics ofpopulation, a smallsample that nearly

represents the whole population is picked and studied. And then the findings are
projected ovu the entire popuhtion to get the approximate overall picture. Sarnpling
methods are basically of two broad tlpes-probability sampling and non-
probability sampling. Again each broad type is divided into subtlpes.
In this unit, you will also leam different sanrpling errors and how to minimize
those errors.

3.1 UNIT OBJECTIVES

After going tlrough this unit, you will be able to:
o Explain the concept of sampling and sampling designs
o Know the steps in sampling design and know the concepts ofuniverse
sampling turit, scores list, size of sample and other matters
o Examine the criteria involved in a sample selection

Self-Instructional Material 69
Sampling Design a Gain fimiliarity with the probability and non-probability sanpling techniques
o Learn how to select right size ofthe sample and how to minimize sampling
EITOTS

NOTES
3.2 SAMPLING DESIGN
Apopulation is the aggregate of all the cases that confofin to the researcher's
designated set of specifications. Therefore, the term people may mean all the
residents oflndia, or those engaged in factory work, or women, boys under the
age of20 and so on, as defined by the researcher. By specification all the boys
under 20 would be included in the population of India, can be referred to as a sub-
population or stratum with reference to the main population.
A stratum may be defined by one or more specifications that divide a
population into mutually exclusive segments, €.g., a given population may be
subdMded into strata ofmales under the age of 21, and females under age of 2l .
Similarly, one can have a stratumbased on education, income, etc.
A single member ofa population is known as an element. Often, one wants
to know how certain characteristics ofthe elements are distributed in a population,
e.g., one wants to the age distribution ofpeople who have a particular political
preference.
A census is a count ofall the elements in a population and a determination of
the distribution oftheir characteristics, based on the information obtained for each
ofthe elements. It is economical in terms of time, effort and money to get the
desired information for only some ofthe elements than for all ofthem.
When we select some of the elements with the intention of finding out
something about the population from which they are taken, we refer to that group
of elements as a sample. The expectation here is that what we find out about the
sanple is true ofthe population as a whole. This depends on the waythe sample
is selected.

Ameasure use based on the entirepopulation is called aparameter. Asample

is any number ofpersons selected to represent the population according to some
rule or plan. So, a sample is a smaller representation ofthe population. Ameasure
based on a sample is known as a statistic.
All items in a field of inquiry are thought to constitute a universe or a
population. Acomplete enumeration ofall the items in the population is called a
census inqutry. It is presumed that a census enquiry gives the highest possible
accuracy. Inpractice, this maynot be so. Even the slightest errorrnayget magnified
as the observations increase. However, such a census tlpe ofenquirycosts lots of
time, money and effort. Therefore, a smaller sample is chosen for study. The methods
used to select samples are called sampling techniques. The surveythat follows is
called a sample survey.

70 Self-lnstuctional Mateial
 Factors influencing decisions while drawing a sanple are: Sampling Design

o Size ofthe population: Whur the population size is large; the selection of
a sample becomes necessary.
o Costs involved in obtaining the elements: Ifthe cost is reasonable; the
sanpling inquiry is fr cilitated. NOTES
. Convenifnce ofavailabilityofthe elements, Each ofthese factors is
mportarfr for deciding to select a sample, for study.
Implications of sample design : Asample is obtained according to a'plan'.
A sanple design is a technique for selecting the items for a sample. The size ofthe
sanple means the number of items to be included in the sample. Sample design
should be determined before data collection and the sample should be designed to
suitthestudy.
Sanpling is a process of selecting a few from a bigger group for estimating
or predicting the prevalence ofsome outcome / factor regarding the bigger groups.
So, a sample is a sub-group ofthe population, one is interested in. This is the
concept ofsanpling see figure 3.1

\ Sample
Population

/*,*.\
obtain from

Vr,,o,._/
Fig. 3.1 Sampling

3.2.1 Steps in Sampling Designing

The followingpoints should be kept in mind before designing a sanple
. Tlpe of universe: Define the universe or set ofobjects to be studied. The
universe can be frrite or infinite. In the finite universe, the numbo ofitems is
certain, while inthe infinite universe, it is uncertain. An exanple ofthe former
is the number ofindustrialists in a country and the latter, the number ofstars
inthe sky.
r sampling unit: This is the group from which the sample is to be drawn.
For exanple, a population unit can be in terms of people's age, gender, etc
and a housing unit lke bungalow, flat, etc, or an educational unit like university,
college, school, etc.
o Service list: The sampling from a list index or other population records
from which the sample is to be drawn, e.g., prepare all the items in the
universe from which the selection ofthe sample can be made. It should be
Self-lnstructional Material 7l
Sampling Design comprehensive, correct and reliable, so that the sample becomes
representative.
o Size of the sample: This refers to the number ifitems to be selected from
the population, to constitute the sanple. An optimum sample size should be
NOTES
reliable, flexible and representative. The size could be determined by the
precision with which estimations are needed. Cost considerations also come
into play, here.
o Parameters of interest: This involves the type ofmeasures needed from
the sample. What factorsyouwantto study.
o Budgetary constraints: This refers to the practical problems about the
size ofthe sample and costs associatedwiththe collection ofdata from the
sample.

3.2.2 Principles for Selecting a Sampling Procedure

There are three principles which guide sampling theory

(i) In majority of cases of sampling there will be a difference between the
sample statistics and the true populationmean, which is attributable to the
selection ofthe units in the sample.
(ii) The greater the sample size; the nure accurate will be the estimate ofthe
truepopulationmean.
(iif The greater the difference in the variable ofthe study in a population for a
given sample size, the greater will be the difference between the sample
statistics and the true population mean.

3.2.3 Systemic Bias and Sampling Errors

There are two reasons for incorrect inferences arising out of sampling. They are:
(i) Systemicbias
(i| Errors in sanrpling procedures
(i) Systemic bias: This bias can arise fromone or more ofthe following reasons:
a. Inappropriate sampling: This means there is a bias in the
representation oftheuniverse, fromwhere the sample is drawn,
b. Defective measuring device: Aphysicalmeasuring device is faulty
or the questionnaire or the interviewer has a bias. This defect would
lead to a systematic bias.
c. Non-responding: Inability to sample all the individuals initially included
in the sample, could give rise to a bias.
d. Indeterminacy principle: Individuals act differently while being
observed. This can cause a systematic bias.
e. Natural bias in the reporting of the data: For example, people

72 Self-lnstructional Material
 tnderstate their incorneswhen the govemment asks forit, but overstate Sampling Design
when social status is involved. In psychological surveys there is a
tendency to give a 'right, answeq rather than a true one.
(ii) Sampling error"rs: These are randomvariations in the sample estimate
around
the true population mean. Sampling errors are effors which arise NOTES
fiom
inaccurate sampling and they generally happen to be random variations
(when sanpling is random) in the sarrple estimates around the
true population
values. see figure 3.2.

Sampling error - Frame error + Chance error * rsspgnss grrsl

(If we add measurement error or the non-sampring i..o, to sampling
error, we get total error)

Fig. 3.2 Sampling Error

source: Kothari, c.R. Research Methodorogy Methods and Techniques, r

995.
Sampling error: Frame error + Chance error + Response error
Sampling error is compensatory in nature. And the expected value of such
error happens to be equal to zero. The magnetite ofthe sampling error
depends
on the nature ofthe universe. Alarger sample design is yet another
wayto reduce
sampling error. Select a sample procedure that helps control systematic
bias and
sanpling error.

3.2.4 Criteria of Good Sample Design

The foilowing arethe criteria ofagood sample design:
(i) Sample design should yield the best possibre sampring error.
(if Sampie design should yield the best possible sampring error.
(iii) Sarnple design shouldbe chosenjudiciously. keeping costs in mind.
Sef-Instructional Mqterial 73
Sampling Design (iv) Sample design must attempt to control systematic bias in the best possible
way.
(v) Sample findings should be applied with a reasonable level ofconfidence.

NOTES 3.2.5 Types of Sampling Designs

For selecting elements on the representationbasis, the sample maybe obtained
using either probability sampling or non-probability sampling. Probability sanpling
is based on random selection whereas non-probability sampling is based on non-
randomsanpling.
On element selection basis, the sample may be either restricted or
unrestricted. When each sample element is drawn individually from the population
at large, the sanple isknownasunrestricted sanple.Allthe otherforms ofsampling
are covered underthe termrestricted sample.

Basic sampling designs

Element selection Representative basis

technique
Probability sampline Non-probability sampling
I Simple random sampling Haphazar d or convenience
Unrestricted sampling sampling
Complex random Purposive sampling, e.g.,
sampling, e.g., cluster quota sampling
Restricted sampling sampling systematic judgment sampling
sampling stratified
san.rpling etc.

Source: Singh, A.K., Tests, Measurements and Research Methods in behavioural sciences,
2m8.

Fig. 3.3 Sampling Designs

Types ofsampling

Simple mndom Stratified random

sampling smpling

Disproportionate
stratified sampling
m-t@
Source: Kothari, C.R. Research Methodology Methods and Techniques, 1995.

Fig. 3.4 Types of Sampling

74 Self-lnstmctional Materiql
3.2.6 Types of Sampling Sampling Design

Strategies for sampling can be categorized as:

o Probability sampling designs
o Non-probability sampling designs NOTES
o Mixed sampling designs

3.2.6.1 Probability (or random) sampling designs

For a random sampling design it is imperative that each element is included in the
sample. Equal element means the probability ofeach element in the population is
the same and is not influenced by other considerations. The notion ofindependence
means that the choice ofone element is not dependent upon the choice ofanother
element in the sample. Only ifboththese conditions are met would the sample be
a random/ probability sampling. If not, bias would be introduced. As this sanple
represents the totalpopulation; the inferences drawn from such samples canbe
generalised to the total population.

Methods of drawing a random sample

The tlreemost commonmethods are
(i) Fishbowldraw method
(u) Using table ofrandom umbers
(ii| Method ofcomputer detennined randomness
The Fishbowl method: This method is useful when the total population
size is small. Here, each element is numbered on a small slip ofpaper and put into
a box or bowl. After the slips are mixed thoroughly, the investigator selects one
number at a time, until the desired sample size is reached.
using the table of random numbers: Most research methodology and
statistics tools have a table ofrandomly generated numbers in their appendices.
The sample can be selected from these tables by following the procedures
described below. The principle ofrandomtable ofnumbers is that the column-
row numbers do not appear in any particular sequence, nor does any number
appear more frequently than the other one.
The random sample selection process:
a. Specify the number ofelements in the study population
b. Number them from 1 to N(//is the total number ofelements in the
po;:ulation). Example, ifthe population size is 500 and one intends to
chcose 50 cases from the population, now enter the random table at
anypoint.
c. Then move systematicallyto the right, left, up, down or diagonally
skipping thenumbers that aretoo large and also thosethat have aheady
been drawn

Self-lnstructional Material 7S
Sampling Design d. Keep moving through the table and fill until the selected sample has
50 elements.

A randomselection of 30 cases frompopulation seriallynrmbered from 1

a
to 80 is illustratedhere.12 numbers have been omitted. Numbers 85, 84, 97 and
NOTES
95 have been omitted because they exceed 80 and numbers 03 , 7 4 and I 2 have
been omitted because they have duplicated the previous selections. The number
00 has been left out because the population number starts from 0l .

The following is a sample random number table :

64755 8388s 84122 25920 17696

10302 52289 77436 34430 38112
710t7 9849s 51308 50374 66s9t
600t2 5s60s 88410 34879 79655

Sample (N: 30 has been taken from a serially numbered population of

80).

64 59 28 12 85 55
75 20 97 7l 03 60
58 t7 74 0l 74
38 69 36 79 66
85 6l 34 84 59
84 03 43 95 l6
t2 02 03 5l 00
22 52 81 30 12

The advantage of using the table of random nurnbers is that it is easily

accessible to the researchers and requires no formal training for using it. However,
the disadvantage is that it cannot be easily and constructively used when the size of
the population exceeds 5 digits.
Method of computer-determined randomness: This method is useful
when the population size is large. The data are fed into the computer to obtain a
random number of elements corresponding to the elements in the population. This
method is easy and fast and therefore is increasingly in use. The only limitation is
imposedbythe need to have a computer.

Different systems of drawing a random sample

There are two ways of selecting a random sample
(i) Sampling without replacement
(u) Sanpling with replacement

76 Self-lnstructional Material
Simple random sample Sampling Design

A simple random sample (also knownas anunrestrictedrandomsample) mayh

defined as one in which each and every individual ofthe population has an equal
chance ofbeing included in the sample and also the selection ofone individual is in
NOTES
no way dependent upon the selection ofanother person. For example, ifwe are to
select a sample of 10 students fromthe seventh grade consisting of40 students,
we can write the name (or roll number) of each ofthe 40 students on separate
slips ofall equal in size and colour and fold them in a similar way. Subsequently,
they may be placed in a box and reshufled thoroughly. Ablindfolded persoq then,
may tre asked to pick up one slip. Here, the probability of each slip being selected
is 1/40. Suppose that after selecting the slip and noting the name writtenonthe
slip, he again returns it to the box. In this case, the probability ofthe second slip
being selected is again 1/40. But ifhe does not return the first slip to the box, the
probability ofthe second slip becomes 1/39. When an element ofthe population is
retumed to the population after being selected it is called sanpling with replacenrent
and when it is not returned, it is called sampling without replacement. Sampling
with replacement is wholly feasible except in certain situations where it is seldom
used (cochran,l963).If sampling withreplacement is used, the chance ofthe
same case being selected more than once is increased. In such a situation, the
repeated cases may be ignored as is done if a table ofrandom numbers is used in
making a selection of cases. Thus random sampling may be defined as one in
which allpossible combinations of samples offixed size have an equal probability
ofbeing selected.
The major difference between sampling with replacement and sampling
without replacement can be illustrated though an example. Suppose the size ofthe
population consists offour persons, who are named as A, B, C and D. Suppose
that the investigator wants to select samples of size 2 through the procedure of
sampling without replacement. In such a situatiorl the investigator can maximally
draw six samples of size 2 fromthe population of4. This could be accomplished
withthe help ofthe following equation.
N N!
(): (N
n -n)lnl
where
N : the size ofparent population
n :the size ofthe sample
! : factorial
In the above example where N : 4 and n : 2, the maximum number of
sample sizeof2 wouldbe 6 as under.

4(- 4t'
u2: 2xlx2xt=244 -^
(4-2)l2l-4x3x2xl
SeA-hslructional Material 77
Sampling Design Similarly, where N:5, we can have 10 sampled of size 2 as under.

tc,: 5!
5x4x3x2xl 120
=lo
(5-2)12! 3x2x1x2xt = I2
NOTES
But fromthe same population, we can have 5 samples ofsize 4 as under:

tco: 5! 5x 4x 3x2xl t20

-\
(5 - 4)!4! lx4x3x2xl 24

But ifthe investigator has decided to proceed with the technique of sampling
-= he can derive the likely number of samples from the given
with replacement,
population withthe help ofthe following equation.
N"
where Nand n are againpopulation and the sample size. Suppose the size ofthe
population is 4 and the size of sample is 2. In such a situation the investigator,
following the technique of sampling with replacement, can maximally draw 16
samples, that is, 42:4x4: 16. Ifthe fourmembers ofpopulation are named asA,
B, C and D, the sixteen samples of size 2 wouldbe
AA AB AC AD
BA BB BC BD
CA CB CC CD
DA DB DC DD

The case ofAA, BB, CC and DD cornbinations reflects the fact that in
sampling withreplacement, an element or individual once drawn canbe drawn
again. In actualpractice such cases are ignored.
There are some advantages and disadvantages of simple random sampling
are as givenbelow:

Advantages of sirrple random sampling:

(i) It is a representative sample.
(i| It is assumed that all the characteristics ofthe population are reflected in the
sample.
(iif This is the easiest and simplest of allprobability sampling methods.
(iv) This random sampling can be applied in conjunction with many other
probability sampling methods.
(v) The sampling error can be easily estimated.

Disadvantages of simple random sampling:

0 It is difficult to ensure that the smaller elements that exist in a population are
included in the sample. For example, in a population of500 persons, only

78 Self-Instructional Material
 12 people are dialectic. The sample size is only 50. The chance that they Sampling Design

would be included is very slim.

(ii) The sirrple random sampling method cannot fully avail ofcertain inforrnation
available in the sample. For example, ifone knows that there is a population
NOTES
ofchildren, who mebright, dull, thosewho are artistic, etc., allthese elements
cannot bebrought into the sample.
(iif Sampling error is greater in the random sample than in a stratified random
sample. To minimize the error stratified random sample is used.

3.2.6.2 Mixed (probability and non-probability) sampling methods

Stratification (also called convenience) sampling method

Stratffied random sampling is of two types
(i) Proportionate stratified random sampling
(ii) Disproportionate stratified random sanpling
(i) Proportionate stratified random sampling
Here the populationis first stratffied. Thenthe sanple is drawnrandomlyfromthe
stratum the proportion ofpersons belonging to each strata ofthe total population
is useful in applied research, e.g., the rank of army officers of one strata in the
defence force and so on. Number ofcardiologists in the totalpopulation ofdarters,
etc., there proportional representation in the sample is a valuable index, for
determining fi.rther studies.
In stratified random sample, the population is first divided into two or more
strata, e.g., onthe basis of age-groups of say 3H0, 40-50, 50-60, etc., or as
male and female. These divided populations are termed sub-populations. These
are non-overlapping parts of the whole population. They are thought to be
homogeneous. Then sample elements are selected from each stratumusing srnple
randomsample.
Advantages of stratified random sampling:
a. In stratified random sampling even those elements that exist in smaller
numbers, get pieced up, and sampling error is minimized because the sample
has all the characteristics ofthe population.

Disadvantages of stratified random sampling:

a. It to start with in terms of assumptions ofknowledge
is a difficult method of
the composition of the population.

b. It is a time consuming method.

c. It is could give rise to classilication error.

d. Disproportionate stratified random sampling.

e. It not
is similar to the earlier me that, but is different in that the substrata are
necessarily distributed according to their proportionate weight in the
Self-lnstructional Material 79
Sampling Design population from which they were randomly selected. There could be
overrepresentation or under representation ofa strata in a population.
(ii) Disproportionate stratified sampling means
NOTES a. The investigator would give equal weight to each ofthe substrata
b. Will give greater or representation to some substrata and not enough weight
to other substrate in the sample to be brawn.
Advantages of disproportionate stratified random sampling:
a. Conparatively lees time consuming than propitiate stratified random sampling

b. It is possible to give weight to particular elements that are not represeated

frequently in the population, as compared to other elements"
Disadvantages of disproportionate stratified random sampling:
a. The samples drawn might suffer from cretin stratum being under or
overrepresented.
b. Assumption that the knowledge ofthe composition ofthe original population
limited use in conditions, where this intbrmation is not available.
c. Possibilityofmisclassrfyingelementsintoasubstrata;thattheydonotbelong,
legitimately, e.g., element Xmaybeput into stratay,but it maybelong to
strataZ.
d. This method is common in the social sciences, despite these limitations.

Area or cluster sampling

This is another inportant method ofprobability sanpling. This is used often in field
research. Here, geographicaldivisions in terms ofterritory zone cities, towns,
districts, etc., are made and a number is assigned to each area. Then the random
sample is drawn from this area. This method is also called cluster sampling. For
example, one wants to study the prevalence ofHIV in a state. Then the state is
divided into districts, downs and villages, etc. identified. Then finafueven households
in localitycould be identified for studying. This is also called multi-stage sampling.

Advantages of area or cluster sampling:

a. This is useful for large scale studies ofarea, regions where specific lists are
an obtainable useful for pubic opinions polls.

b. It saves time and money

c. The respondents can easilybe substituted. It is a flexible method
Disadvantages of area or cluster sampling:
a. Sampling errors rarelycome to light.
b. Little control over even the size ofthe cluster
c. Could result in bias in samples
However this method is used in large behaviour studies.
80 Se{-lnstnrctional Mateial
3.2.6.3 Non-probability sampling methods Sampling Design

Quota sampling
This is an important non-probability sampling method. The population is seen as
made up ofstrata ofthe population and from each stratunr, individuals are chosen NOTES
randomly, e.g., ifthe population ofstudents in a school is 5000 made up ofhigh
and low socio-economic classes. From this, 500 students can be chosen with 250
fromthe higherand 250 fromthe lower class. This is the quota sample.
Advantages of the quota sampling method:

@ Quota sampling method is quick and easyfor gross estimates

(ii) It is a convenient procedure
(ii! Elements fromthe desired strata get included
Disadvantages of the quota sampling method:
(i) In quota sampling method randomness cannot be established

(i| In this method, generulaability is poor

(iir) since most credible elernents included, it maynot be tlpical ofthe population
(iv) Possibility ofclassification error high.
(v) Other variables ofsignificance cannot be controlled.

This method does have some appeal, despite limitations.

Purposive sample
Purposive sample, which is a handpicked sample is typicalofthe population. lt is
also calledjudgmental sample, because the choice is determined bythe judgments
of the researcher, e.g., attitudes towards comrption would be ascertained by
interviewing professionals, academicians, tainted people and politicians. The
investigator selects the persons from these select people.
Advantages of purposive sample:
(i) Purposive sample is cost effective and easily accessible

(if It isveryconvenient
(ri| Onlyrelevant individuals get included
Disadvantages of purposive sample:
(i) Inpurposive sampling randomness representatives not ensured.
(if Generalizability is poor.
(ii| Sampling is highly subjective.
(iv) Inability to apply inferential statistics to acceptable levels
Accidental sampling
Accidental sampling is also called incidental sampling. It is based on a non-
probability sampling plan. Here, the investigator chooses the sample according to
Self-lnstructional Material 81
Sampling Design convenience. Convenience and economy guide this method, as a useful option,
e.9., students ofa particular school chooses because ofavailability

Advantages of accidental sampling:

NOTES (i) Accidental sampling is the most convenient method ofsampling
(i| Economical in terms oftime, money, effort and cost.
Disadvantages of accidental sampling:
(i) Accidental sampling has poor generalizability

(i) Biases and prejudices bound to exist.

(iif There is a high probability ofhigh sampling error
Accidental sampling is used inpsycho-social research due to the convenience
factor.

Systematic sampling
This is another method ofthe non-probability sarrpling plan. Here, every rth person
is drawn from a predetermined list for study, e.g., every 5th roll number ofclass of
50 students or every 1Oth name from the telephone directory and so on. It is
systematic in view of the fact that the selection is made according to a pre-
determined plan. The first element selected is random and has non-probability
characteristics

Advantages of systematic sampling:

(i) The method is quick and it is easyto obtain the sample.

(D It is easy to select, say, every fifth name.

(rlD It is easyto use (simpler thanrandomtable chart)
Disadvantages of systematic sampling:
(i) It is not a probability sampling plan
(ii) Hlgh sampling errorpossibilitycreeps in
(ii) Bias could invade into the sample
This method still finds favour inpsycho-social research.

Snow ball sampling

Snow ball sampling is a non-probability sampling method. It is basically a socio-
metric method. Here all the persons ofa group or sample are identified by friends
or other acquaintances. The snowball effect is how one person's contact leads to
the othersandfromthere furtheron. The informationgatheredisthroughpatterns
offriendship, e.g., how rumors spread and how advertisements influence friends,
etc. It is useful for small N samples, below 100. It is a method to study social
change.

82 Self-lnstructional Material
Advantages of snow ball sampling Sampling Design

(i) Snow ball sampling is a systematic sampling technique helpful in studying

small informal networked social groups.
(if Helps in community studies, decision-making
NOTES
(iit) This technique lends itselfto computer-determined use ofrandom numbers
Disadvantages of snow ball sampling
(i) Snow ball sampling method is not feasible with large numbers
(0 No use ofstatisticalmethods, possible
(iir) Bias could enter the sample
(iv) Largely usefirl for ffirmation dissemination, saturation sampling and dense
sampling.

Saturation sampling
Saturation sampling involves drawing all elements or individuals having
characteristics that are ofinterest to the researcheq e.g., allpsychiatrists below the
age of45 years. Dense sampling is a method that lies between simple random
sanpling and saturation sampling. Here researcher selects 50 per cent or more
from the popirlation and takes a rnajority ofindividuals having specific characteristics
that are of significance, e.g., 500-600 students from a class of 1000 students.
These two methods are convenient. But it is not useful when the Nexceeds 1000.

Double sampling
Double sampling means drawing a sample of individuals from a sample that has
already been drawn earlier, e. g., from a population of 1 0, 000 people, a sampling
of 1000 is drawn. Again fromthis 1000, a further sample 200 is drawn, for the
study, e.9., o questionnaire is sent to I 000 people on the issue of pollution. Say,
50 per cent (or 500) ofthem respond. From these 500 persons, a sample of I 00
is drawn for an in-depth interview. This is double sampling.

Ifthis method is to be meaningful for research, thenthe representativeness

ofthe sanple must be increased It is a time-consuming and labour intensive method.
Pre-requisites for a sample to be good are as follows:
(i) It must be representative
(i| It must be ofsufficient size
(i) Representativeness: The sample must have approximately all the
characteristics ofthepopulationthat are relevant forthe investigation. The
relevant characteristics ofthe population must exist in the same proportion
in the sample also, e.g., ifthe population has an education level ratio of
60:40 for men and women, the sample should also have a similar ratio, to
be representative. The population should be clearly defined and the
observations to be made should be specified. This can reduce biases.

Sef-Instructional Material 83
Sampling Design (ii) The sample size should be adequate: This implies that the size be
zufficient. Alarger sample is better forreducing the error. This is the difference
between the population value and the sample value. The larger the size of
the sample, the lesser is the error. However, too large a sample maynot
NOTES yield better results, as a large sample creates other problems.
Advantages of sampling methods in general
(i) Using sampling methods increases accuracy. Examining a sample becomes
both efficient and involves lesser work, so the purpose ofa sample is to get
nriximum accuracywith minirnal effort, time, money, etc.
(ir) It reduces the cost as the data is from a smaller number ofcases. Statistical
calculations for accidental errors are also reduced.
(iif Since the sample and not the udverse is studied, workproceeds fast. This
is a great advantage for research.

CrmcxYouR PRocRESS

1. What is a census survey?

2. What is a simple random sample?
J. Why are probability sampling procedures preferred to non-probability
sampling procedures?
4. What israndomsampling?
5. Whatisquotasampling?
6. What is systemic bias?
7 . Define cluster sampling.

3.3 STATISTICAL DISTRIBUTIONS

The following three t1,pes of distributions are of interest to any researcher:
(i) Sample distribution
(i| Population distribution
(rir) Sarrpling distrbution

Fundamentals of sampling
Sampling is the process of obtaining information about an active population by
examining only a part of it. one draws inferences about the parameters of a
population from which the samples are obtained. The assumption behind this is
that the sanple data would enable an estimate ofthe population The items selected

E4 Sef-lnstructional Material
from the population for observation is called a sample. The method of selecting a Sampling Design
sample is called sample design. Any survey that is carried out on the basis ofthe
sample is referred to as sample survey. To draw valid and reliable conclusions, the
sample must be truly be representative ofthe population.
NOTES
Need for sampling
Sampling reduces time and money. It is less expensive and requires less time.
Sampling makes measurements more accurate, due to smaller size.

Sample is the onlymeans to studya large population. when items are to be

deconstructed, sample is the onlymeans for study.
Some fundamental definitions: These are to familiarize oneselfwiththe concepts
ofsampling.
(i) universe population: The toral number of items in any field of study is
called the universe. The population refers to the total units or items about
which information is required. The attributes that are the object ofthe study
are called the characteristics and the units possessing them are known as
elementaryunits. The aggregate ofsuch units is the population.
All urits in any field of study constitute the turiverse. All elementary-units are
the population often the two terms are used interchangeably, however,
research needs this distinction. The population or universe can be of two
types: (i) Finite and (ii) Infinite.
A finite population consists of fixed number of elements and the elements
can be enumerated totally, e.g., the nurnber ofstudents in a state. The symbol
Nis used to depict the number of elements or items ofa finite population.
An infinite is one where all the elements caru1ot be observed, at least
theoretically, e.g., the number of stars in the sky. In a sense, a very large
finite population is an infnite population.
(ii) sampling frame: The elementaryunits that form the basis ofthe sampling
process are known as sampling units. A list of all such sampling units is the
sampling frame. The sampling frame is a count ofa rist ofitems from which
the sample is drawn. For research, a frame has to be constructed from the
population, to draw the sample, e.g., names from the census records or
telephone directory etc., for conducting a study on a sample that is drawn
from the fiame. Telephone directory is a franre, from which names are drawn
to get the sample.
(iif sampting design: This is the plan for obtaining a sample fromthe frame. It
is theprocedwe or technique for obtaining those sampling units from which
inferences can be made. The sampling design has to be prepared well in
advance ofundertaking any research.

Self-Instructional Material 85
Sampling Design (iv) Statistic(s) and parameter(s): Astatistic is the characteristic ofthe sanple
whereas the parameter is the characteristic ofthe population. Sampling
analysis involves estirnating the parameter fromthe statistic.

(v) Sampfing er:ror: This inforrnation collected because

is the inaccuracy in the
NOTES
only a small portion ofthe population is inchrded in the study. The sampling
error is also known as errorvariance. Sampling erors arise out of sampling
and are generally random variations in the sample estimates around the true
populationvalues.
Sampling elTor: Frame error * Chance error + Response error
Sampling errors decrease as the homogeny of the universe increases.
Sampling error is usuallyworked out as the product ofthe article value at a
certain level of significance and the standard error.
Non-sampling effors also occur during the process ofthe data collection.
But these cannot be measured. Since the sample and not the universe is
studied, work proceeds fast. This is a large advantage for research.
(vi) Sample distribution: For example, say, from a population of30,000, a
random of300 persons is chosen for a given study. The observed data are
arranged ina frequency distnbutioq e.g., fertiliryrate. This type ofdistribution
is called a sample distribution. This is the data got from a sample ofthe
population.
(vii) Population distribution: Ifthe fertilityrates ofallthe 30,000people ofthe
population are obtained and arranged in a frequency distribution, it is known
as the population distribution. Since the forms and parameters are not
ordinarily known, an estimate ofthese two characteristics ofthe population
is made from the sanple distribution. So, ifthe sample distribution is normal
one can assume that the population distribution is also normal.
(vii) Sampling distributions: Inportant sanpling distnbutions that are commonly
used are:
a. Sanplingdistributionofmean
b. Sampling distnbution ofpopulation
c. Student's r distnbution
d. Fdistnbution
e. Chi-square distribution
a. Sampling'distribution of mean: This refers to the probabilily distribution
ofall possible means ofrandom samples of a given size taken from the
population. If samples are taken from a normal population, then the
sampling distribution ofmean would also be norrnal But when the sanple
is not from a norrnal population (a skewed one negatively or positively),
even then as per the central limit theorem, the sampling distribution of
mean tends to be closer to the normal distribution, providedthe sample
size is large--more than 30.
86 Self-lnstructional Material
b. Sampling distribution ofproportion: One can work out the proportion Sampling Design

of items from a population, e.g., boys under the age of20 in a town.
This would give the sampling distribution ofproportion ofboys under
20 to the whole population. As the size increases; the distribution tends
to become normal. NOTES
c. Student's / distribution when the population standard deviation is not
known and the sample size is smaller than 30, the I distribution is used.

t:

(,
s

Here, the sample standard deviation I distribution is also symmetrical

and is veryclose to the distribution of standard normal variate, z, except
for small values of n. The variable / differs from z in the sense that one
uses sample standard deviation (o ) while calculating I whereas standard
deviation ofpopulation (o ) is used for c alubtmgz. This gives a different
I distribution for eachpossible sample size, that is, for different d.f The
d.f for a sample of size n is n l. As the sample size increases, the
-
shape ofthe r distribution becomes approximately equal to the normal
distribution. In fact, if the sizes are more than 30, the / distribution
approaches the normal distribution so close that we can use the normal
distribution to approximate the I distribution. But when n is smal1, the r
distribution would be far from normal distribution, while when n : a,, t
distribution is identicalwithnormal distribution. The I distributiontables
show the critical values of r for different d.f. at various levels of
significance. The table value of I for given d.f. at a particular level of
significance is compared with the values of r obtained from the sample
data, and ifthe latter is either equal to or exceeds the former, one can
infer that the null hypothesis cannot be accepted.

d. F distributio,n: If(cr.,)2 and (cr.r)2are the variances oftwo independent

samples of sizes n, andn, respectively taken from two independent
normalpopulations, having the same variance, 1oo), : (oor).,the ratio
F: (q,), (orr),, where (o,,)r: I(t,, - X)rln, - 1 and (o-r)2
:L(Xr,- Xr)'lnr- I has anFdistribution with n,- | andnr- I
degree of freedom. Fratio is computed in such away that the larger
variance is always inthe numerator. Tables forFdistributionthat are
available give criticalvalues ofFforvarious values ofdegree of freedom
for larger as well as smaller variances. The value ofFobtained fromthe
sample data is compared with the corresponding table value ofF and if
the former is equal to or exceeds the latter, then the inference drawn is

Self-lnstructional Materiol 87
Sampling Design that the null hlpothesis ofthe variances being equal cannot be accepted.
The Fratio canbe used in the context ofhypothesis testing and also in
the context ofANOVA t echnique.

NOTES e. Chi-square (f ) distribution: Ifeach collection ofsample

variances is
divided by the known population variance and these quotients are
multipliedby(n - 1) (n stands forthe number ofelements inthe sample),
the chi-square distribution is obtained. Thus, 1o (n * l) would
"2lor)2
have the same distribution as chi-square distribution with d.f. as (n -l).
Chi-square distribution is not symmetrical with all the values being
positive. For using chi-square distribution, knowledge ofthe degrees of
freedom is necessary.

Distribution is also used for judging the significance ofdifference between

the observed an{ the expected frequencies and also as a test ofthe
correctness ofthe fit. The generalized shape of 12 distribution depends
upon the d.f and the y2 value is worked out in the follorving manner'

l\
^,.:
" +(o,-8,)2
2
r=l Ei

There are tables available that give the value of x' fbr given d.f. and these
can be used for calculating values of 1' for relevant d.f, at a desired level of
significance for testing hlpotheses.

Central limit theorem

When a sample drawn from a population which is not normal, the sample size
plays a criticalrole. Ifthe populationis normal, the means ofsamples drawn from
ihat populationwouldthemselvesbe normaliydistributed, but when sampling is
not from a normalpopulation, the sample size plays a signifrcant role. When n is
small, the shape ofthe distributionto a great extent depends onthe shape ofthe
parent population, but as n becomes large (r > 30), the shape ofthe sampling
distribution will move closer to norrnal distribution. inespective ofthe shape ofthe
parent population Centrallimit Theoremin deed explains this relationship between
the shape ofthe population distnbution and the sampling distribution ofthe mean.
Central limit theorem has statistical inference, as it expalins
a great significance in
how the sampling distribution ofthe means approaches normal distribution with
the increasing sample size. In technicalterms, the distnbution ofnreans ofrandom
samples from a population with mean pr and finite variance o' approaches the
normal distribution with mean p and varianc e o'ln as r tends to infinity.
The centrallimit theorem's significance is that it permits the use of sample
statistics to make inferences about population paranreters without knowing anything
about the shape ofthe tequency distribution ofthat population other than what is
available from the samPle.

Se lf- I nstruct io na I Ma te ri a I
 Sampling theory Sampling Design

S ampling theory deals with the relationships between a population and random
samples drawn from the population. population or a univeise is an aggregate
of
items with common traits. A universe constitutes the totalilty ofthe items
about
which researcher seeks study. The universe may be finite or infinite. Finite universe NOTES
contains a definite number or items. In an infinite unverse the number ofitems is
infinite.

The universe may be hypothetical or real. In the hypothetical case, the

universe does not exist and one only imagines the items constituting it. Tossing
a
coin and throwing the dice can be cited as examples ofa hypothetical universe.
The real universe consists ofconcrete objects. Sample is thatpart ofthe universe
which is selected at random for the purpose of investigation.
sampling theorymainlydeals with the relationship between a parameter
and a statistic. The theory estimates the properties ofthe population on the basis
ofthe sample and also etaluates the precision of the estimaie. This is known as
statistical induction or statistical inference as it attempts to draw the inference
conceming the universe from the sample. To use this inductive method, first follow
a deductive argument-imagine a universe (finite or infinite) and investigate
the
behaviour ofthe sarrples drawn fromthis universe applyrng the laws ofprobability.
Such methodolergyis known as sampling theory The theory's objectives are listed
below:
(i) statistical estimatioz.' This consists of estimating the unknown
population parameters from a knowledge ofstatistical measures based
on sample studies. The estimate can either be a point estimate or an
interval estimate. Point estimate is a single estimate expressed in the
formofa single figure. while intenal estimate has two limits, viz.,the
upper limit and the lower limit within which the parameter value rnay
lie. Interval estimates are the ones which are often used in statistical
induction.
(r\ statistical inference: The sampling theory helps in arriving at
generalizations about the population / universe from the studies based
on samples drawn from it. It also helps in testing the accuracy ofsuch
generalizations.
The theoryofsanpling can also be studied trnder two lreads: (D The sampling
ofattributes and (ii) The sampling ofvariables in the context oflarge anA smatt
samples. Asmall sample is one that has 30 items or less whereas u lurg. sample
has more than 30 items. When studying the qualitative characteristics
ofall items in
apopulation' trne obtains statistics ofattributes in the form oftwo classes----one
in
which the attnbute is present and the second where it is absent. The presence
of
an attribute may be termed as a 'success' and its absence a .failure,.

The theory can also be applied in the context of statistics ofvariables (i.e.,
data relating to some characteristic conceming population which can be estimaied).
The objectives are:
Self-Instructional Material 89
Sampling Design . To compare the observed and expected values ofthe sample and to
determine if the difference can be ascribed to the fluctuations of
sampling;
o To estimate the population parameters fromthe sample, and
NOTES
o To find out the degree ofreliabilityofthe estimate.
The tests of significance used in dealing with problems arising in studying
large samples are different from those used for small samples. This is because the
assumptions that one has to make in the case of large samples do not hold good
for srnall sarrples. It is assumed in case oflarge sanples that the sarrpling distribution
tends to be normal and the sanple values are approxirnately close to the population
values. This helps in applyrrg what is known as the z-test. When n is large, the
probabilityofa sample value ofthe statistic deviating fromtheparameterbymore
than three times its standard error is very small (it is 0.0027 as per table giving
area under normal curve). The z-test, thus is applied to find out the degree of
reliability ofa statistic in case oflarge samples. One, ofcourse, needs to work out
appropriate standard errors as they will enable one to give the limits within which
tlreparametervalues would lie orwouldenable one to judgewhetherthe difference
happens to be significant or not. For example, X + 3o, would give the range
within which the parameter mean value is expected to vary with 99.73 per cent
confidence level.
The sampling theory that is applied for large samples is not applicable in the
case ofsmall samples because in the case of samples, one cannot assume that the
sampling distributionis approximatelynormal.Adifferenttechnique is required for
handling smallsanples inparticularwhen thepopulationparameters areunknown.
Sir William S. Gosset developed a significance test, known as student's I test,
based on / distribution. His was a significant contribution to the theoryofsampling,
applicable in case ofsrnall samples. Student's / test is used whentwo conditions
are fulfilled: the sample size is 30 or less andthe population variance is not known.
While using I test, one assumes that in the population from which the sample has
beendrawn:
o The sampleisrandomlydrawn
o Observations are independent
o There is no measurement error
o And that in the case oftwo samples where equality ofthe two population
means is to be tested, one assumes that the population variances are equal.

To applythe /test, one needs to work out the value oftest statistic (i.e., r)
and then compare it with the table value of / (based on / distribution) at certain
level of significance for given degrees of freedom. Ifthe calculated value of r is
either equal to or exceeds the table value, the inference is that the difference is
significant. But ifthe calculated value ofr is less than the conceming table value of

90 Self-lnstnlctional Material
l, the difference is not treated as significant. The following formulae are commonly Sampling Design
used to calculate the rvalue:
(r) To test the significance ofthe mean ofa random sample:

(x-rr) NOTES
t-
or

where X :Meanofthe sample

p : Mean ofthe universe / population
o : Standard error ofmean worked out as under

and the d.f. : (z _1).

(i} To test the difference between the means oftwo samples
x,- R,
I_ or,_.,

where X, :Meanofsampleone
X, : Meanof sample two
c *,- N, : Standard error of difference between two

Sample means worked out as

6_-:
xt -xr

And the d.f. = (n, + nr-2).

(nD To test the significance ofthe coefficient ofsimple correlation

t' : -lxJna or t=r^E

\l-r'
Jl-r'
where
r:
thecoefficient of simple correlation anC the d.f.: (n
-2).
(iv) To test the significance ofthe coefficient ofpartial correlation
ro 4)
t:--xln-K Ort=r.,\t'fr '
,lr_,, r_,;
whele r is any partial coefficient of correlation and the d. f : (n k),
o -
z being thenurnber ofpairs ofobservations andtbeingthe number of
variables involved.

Self-Instructional Material 9I
Sampling Design (v) To test the difference in case ofpaired or colrelated samples data (in
which case I test is often described as difference test).

D r-
-lto tln;o D-o
t--ln
r-
NOTES
t -- oD

where hypothesised mean difference (;.lr) is taken as zero (0),

D : Meanofthe differences ofcorrelated sample items.

or: Standard deviation ofdifferences worked out as under.

l>ol - o,'n
o': 1/ a-1
D. = Differences { i.e., D.: (X;Y)}
n : number ofpairs in two samples and the d.f. : (n - 1)

Sandler's I test
Joseph Sandlerhas developed an altemate approach based on a simplifiedversion
of r-test. Sandler's A test that serves the same purpose as is accomplished by
/-test relating to paired data. Researchers can also use l-test when correlated
samples are employed and hypothesized mean difference is taken as zero, i.e.,
Hn: pr: 0. Psychologists generally use this test in case oftwo groups that are
matched with respect to some extraneous variable(s). While using l-test, one has
to work out l-statistic that yields exactly the same results as student's /-test.
A-statistic is found as follows.

the sum of squares of the differences

- the square of the sum of the differences
=L4
(x4)'
In altest, the d.f. is the same as in Student's / test, i.e., d.f. : fi - l,n being
equal to the number ofpairs. The critical value of,4, at a given level ofsignificance
for given d.f , is possible to obtain from the table of A statistic.One needs to
compare the computed value ofl with its corresponding table value for drawing
an inference as to the acceptance or rejection ofnull fupothesis. Ifthe calculated
value ofl is equalto or less thanthe table valuethel-tatistic is considered significant
and thereupon one can reject Hoandaccept F1,. But ifthe calculated values ofl
is more than its table value, thenA-statistic is taken as insignificant and accordingly
I1o is accepted. This is because the two test statistic s, viz., t and A are inversely
related. These two statistics can be written in terms ofone another in the following
lnamer:
(i) ,,4 in term of I can be expressed as

n-l I
,,|I _
- n.t' n
-L-

92 Self-lnstructional Mlterial
 (i) I interms of A canbe expressed as Sampling Design

t-
NOTES
Thus, computingl-statistic is relatively srnple. Using ofl-statistic heSs in
saving time and labour considerabty, especiallywhen the matched groups are to
be compared with respect to a large number ofvariables. Researchers would be
well advised to replace student's I test by Sandler's ,,4 test whenever correlated
sets of scores are employed. Sandler's l-statistic can also be used 'in the one
sample case as a direct substitute for the student t-ratio.' because Sandler's I is
an algebraically equivalent to the Sfudent's /. when one uses l-test in one sample
case, the following steps are involved:

(i) Subtract the hypothesised mean ofthe population (zr) from each
individual score (,q) to obtainD. and thenwork out ID
(u) Square each D and then obtain the sum of such squares i.e., D!
(O Find,4 statistic asunder:
tr:fi21(LD,)2
(iv) Look up the table of A statistic for (n - 1) degrees of freedom at a
given level of significance (using one-tailed or two-tailed values
depending upon H") to find the critical valu e of A.
(v) Finally, the inference canbe drawnas indicatedbelow:
When calculated value ofl is equal to or less than the table value,
rject ,I1o (or accept 4) but when computed A is greater than its table
value, then accept .Flo.

Concept of standard error

The standard deviation of the sampling distribution ofa statistic is known as its
standard enor (S.E.) and is considered the key in sampling theory. The utilityof
the concept ofstandard error in statistical induction lay in:
(i) The standard error helps in testing whether the difference between the
observed and the expected frequencies would arise due to chance. The
usual criterion followed is to find out if a difference is less than three times
the S.E. The difference is supposed to arise as a matter ofchance. Ifthe
difference is equal to or more then three times the S.E. , the chance fails to
account for it, and the conclusion drawn is that it is a significant difference.
The criterion is based on the fact that at + 3 (S.E.), the normal curve
covers an area of99.73 per cent. Sometimes the criterion of 2 S.E. is also
used in place of 3 S.E. Thus the standard error is an important measure in
signfficance tests or in examining hypotheses. Ifthe estimated parameter
differs fromthe calculated statistic bymore than 1.96 times the S.8., the
difference is takenas significant at 5 per cent level ofsignificance. In other

Self-lnstnrctional Material 93
Sampling Design words, the difference is outside the limits, i.e., it lies in the 5 per cent area
(2.5 per cent onboth sides) outside the 95 per cent area ofthe sampling
distribution. Hence one can conclude with 95 per cent confidence that the
said difference is not due to fluctuations of sampling. In such a case, the
NOTES
hypothesis that there is no difference is rejected at 5 per cent level of
significance.
But ifthe difference is less than 1.98 times the S.E., then it is considered not
significant at 5 per cent level. It can then be said with 95 per cent confidence
that it is because ofthe fluctuations of sampling. In such a case, the null
hypothesis stands true. 1.96 is the critical value at 5 per cent leve!. The
product ofthe critical value at a certain level of significance and S.E. is
described as 'sampling error'at that particular level ofsignificance. One
can test the difference at certain other levels of significance as well depending
upon one's requirement.

Criteria for Judging Signrficance at Various Important Levels

Significance Confidence Critical Sampling Confidence Difference Difference

level level value error limits Significant Insignificant
if if
5.0% 95.0o/o 1.96 1.96o +1.96o ! 1.96o ? 1.960
1.0% 99.0% 2.s758 2.5758o +2.57580, i 2.5758o 72.5758o
2.7o/o 99.73% 3 3o +3o !3o ?30
4.55o/" 95.45% 2 2o +2o )2o Z2o

(D The S.E.also provides a good measure

ofreliability and precision ofa sample.
The smaller the S.E., the greater the uniformity of sampling distribution and,
therefore, greater is the reliability ofthe sample. Conversely, the greater the
S.8., the greater is the difference between observed and expected
frequencies. In such a case, the unreliabilityofthe sample is higher. The size
ofS.E., dependsuponthe sample sizeto almge extent anditvaries inversely
with the size ofthe sanple. Ifreliabilify factor is to be doubled i.e., reducing
S.E. to 1/2 ofits existing magnitude, the sample size should be increased
four-fold.
(o The S.E. lso enables one to specifythe limitswithinwhichthe parameters of
the population with a specified degree of confidence. Such an interval is
known as confidence interval. The following table gives the percentage of
samples having their mean values within arange ofpopulation mean (p) +
S.E.

Range Per cent Values

p +l S.E. 68.27%
p +2 S.E. 95.45%
p +3 S.E. 99.73%
p +1.96 S.E. 95.00%
p +2.5758 S.E. 99.00%

94 Self-lnstuctional Material
Important formulae forcomputingthe standard errorsconcemingvariousmeasures Sampling Design
based on samples are as under:

a. In case of sampling ofattributes

i Standard error ofnumber of successes : n'p'q NOTES

where
n : number ofevents inn each sample,
p : probability of success in each event,
q : probability offailure in each event,

ii. Standard error ofproportion of successes : p'q

n
iii Standard error ofthe difference between proportions oftwo samples:
wherep: best estimate ofproportion ofthe population and is worked out
as under:

r)(r
p: p'ql\.u,
-+-nz ) I

q:l-q
z, : number ofevents in sample one
n, : nurnber ofevents in sample two
Nale.'Instehd ofthe above formula, we use the following formula:

Pflt , PzQz
6 pr-p, :
t\ n2

when samples are drawn from two heterogeneous populations and where we cannot
have the best estimate of proportion in the universe on the basis of the given sample
data. Such a situation often arises in the study ofassociation ofattributes.

b. Incase ofsampling ofvariables (large samples):

i Standard error ofmean when population standard deviation is known:

C,
o-: /-
1n
where
o_
p
: standard deviation ofpopulation
r :numberofitems inthe sample
Note: This formula is used even when r is 30 or less.

Self-Instructional Material 95
Sampling Design IL Standard effor ofmean population standard deviation is unknown:

os
o-:-F
" ^l
n
NOTES
where
o,2 : standard deviation ofthe sample and is worked out as under

>(x,- N)',

: number ofitems in the sample

iii Standard error of standard deviation when population standard deviation is

known:
co
o",:
J2n
iv. Standard error of standard deviation when population standard deviation s
unlcnwn:
or
o".: ,f 2n

>(X, - N)'
0,:
,-t
where
n : number ofitems in the sample.
v Standard error ofthe coefficient ofsimple correlation:

I- 12
o: _G
where
r: coefficient ofsimple correlation
n :numberofitems inthe sample
vi. Standard error ofdifference betweenmeans oftwo samples:
c. When two samples are drawn from the same population:

0t,-.", =

(If oo is not known, sample standard deviation for combined samples

(q,r) may be substituted).

96 Se{Jret'uctional Material
 d. Whentwo samples are drawn fromdifferent populations: Sampling Design

or-,-x, =
NOTES
llforr and or.rare not known, then in their places o., and orrrespectively
maybe substituted.)
e. in case of sampling ofvariables (small samples) :

i Standard error ofmean when or ls unknown:

n,(a + nr(o,.)z + n,(X,

- X,,r)' + nr(X, - X,,).
Q",.: {lI ",)2
'' \l \+n2
. * - nttx,)+ nr1X.)
where t.z
A-,^ nt + n2
Note: (l) All the above formulae apply only for infinite population. In cse the population is
finte, sampling is done without replacement and the sample size is more tharr5 per cent of
the population, the finite population multiplier must be used in the S.E formulae. For example,
SE_ for finite population will be as shown below:

6p
sEx: Ji
In cases in which the population is very large in relation to the size ofthe sample, the finite
population multiplier is close to one and has little effect on the calculation of S.E. In such a
case where the sampling fractiotl is less than 0.05, the finite population multiplier is not
generally used.

: *= n-1
".r !n J,
ii standard error ofdifference between two sample rneans whor o, is unknown

6--=
.\t- A t
E=
!', n2

Estimatin g population parameterc

In most statistical research studies, population parameters are usually not known
and therefore, have to be estimated from a sample. Here, methods for estimating
the populationparameters have an important role in statistical anatysis.

The random variables (such as and q') used to estimate population:

conventionally, parameters such as p and o.2 are called as .estimators,,
while specific values of these (such as : 105 or o-2 : 21.44) are referred to as
'estimates' ofthe population parameters. The estimate ofa population parameter
may h one single value or a range ofvalues. In case of single value parameter, it

Se{-Instructional Moterial 97
Sampling Design is referredto as 'point estimate,' and in case where there is a range ofparameters,
it is termed as interval estimate. The researcher has to make these two types of
estimates through sampling analysis. While making estimates of population
parameters, the researcher can give only the best point estimate as otherwise
NOTES he/she has to speak in terms of intervals and probabilities; For, he/she can never
estimate with certainty the exact values ofpopulation parameters.
A good estimator possesses the following properties:
(1) An estimator should on the average be equal to the value of the
parameter being estimated. This is known as the property of
unbiasedness. An estimator is said to be unbiased if the expected
value ofthe estimator is equalto the parameterbeing estimated. The
sanple mean 1X; is the most widelyused estimator because it provides
an unbiased estirnate ofthe population mean (p).
(i} An estimator should have a relatively small variance.This property is
technically described as the property ofefficiencybecause an estirnator
is expected to have the smallest variance.
(trD An estimator is expected to use as much informatrion as possible as
is available from the sample. This property is known as the property
ofsufficiency.
(iv) As the sample size becomes larger and larger, an estimator has to
approach the value ofpopulation parameter. This property is refened
to as the property ofconsistency.
The researchermust select the appropriate estimator(s) forhiVher studyby
keeping in view the above stated properties.

Estimating the population mean (p)

Inpoint estimate, the sample mean )( is the best estimator ofthe population mean,
m and its sampling distribution when the sample is sufficiently large, because this
tends to approxirnate the normal distribution. Ifone knows the sampling distribution
of X , one can make statements about any estimate that one may make from the
sampling ffirmation. In a sanple of36 students, ifone finds that the sample yields
an arithmetic mean of 6.2 i.e., X : 6.2, replace these student names on the
population list and draw another sample of36 randornlyand assume that one gets
a mean of 7.5 this time. Similarly a third sample mayyield a mean of6.9; fourth a
mean of6.7; and so on. One can go on drawing such samples till one accumulates
a large number ofmeans of samples of 36. Each such sample mean is a separate
point estimate of the population mean. When such means are presented in the
form of a distribution, the distribution happens to be quite close to normal. This is
acharacteristic ofadistnbutionofsanplemeans (andalso ofothersanple statistics).
Even ifthepopulation is not normal the sanple means drawn fromthat population
are dispersed around the parameter in a distribution that is generally close to
normal; the mean of the distribution of sample means is equal to the population

98 Self-Ins*uctional Material
 mean. This is true in case of large samples as expected from the central limit Sampling Design
theorem. The relationship bet'ween a population distribution and a distribution of
sample mean is critical for drawing inferences about parameters.

The relationship between the dispersion of a population distribution and

that ofthe sample meancanbe expressed inthe following formula: NOTES

co
O-: --F
" ,l n

where o : standard error ofmean ofa given sample size

o, : standard deviation ofthe population
n :sizeofthe sample.
How to find o when one has the sample data only for one's analysis? The
answer is that one n{ust use the best estimate of o- and the best estimate can be
the standard deviation ofthe sample, o . Thus, the itandard error ofmean can be
worked out as under:

os
-, ---
6-:
,l n

>(X,- X)'
where o:s
n-l
With the help ofthe above, one can arrive at intervalestimates about the
parameter inprobabilistic terms (utilising the fundamentalcharacteristics ofthe
normal distribution). Suppose one takes a sample of 36 items and works out its
mean 0 to be equal to 6.20 and its standard deviation (q) to be equal to 3.g:
Then the best pgint estimate ofpopulation mean (p) is 6.20. The S.E ofmean (o)
would be 3.8 / {36 : 3.816:0.663. And ifone takes the interval estimate of p to
ue x t 1.96 ( o';; or 6.20 tr.24or from4.96 to 7.44,it means that there is a 95
per cent chance that the population mean is within 4.96 to 7.44 ntewal. This
means that ifone were to take a complete census ofall items in the population, the
chances are 95 to 5 that one would find the population mean between 4.96 to
7 .44.In case one wants to have an estimate that will hold for a much smaller

range, then one must either accept a smaller degree ofconfidence in the results or
take asample large enoughto provide this smallerintervalwithadequate confidence
levels. Usually one thinks of increasing the sample size till one can secure the
desired interval estimate and the degree ofconfidence.
Illustration 3.1 : From a random sanple of36 New Delhi civil service persorurel
mean age andthe sample standard deviationwere found to be 40 years and 4.5
years respectively. Construct a 95 per cent confidence interval for the mean age of
civil servants in New Delhi.

S elf- In struc lio na I Moterial

Sampling Design Solution: The given information can be written as under:
n :36
x:40years
NOTES o. :4.5 years
And the standard variate, z, for 95 per cent confidence is 1.96 (as per the
normal curve area table).
Thus, 95 per cent confidence interval for the mean age ofpopulation is:

X tr9+
Jn

61 4o+ r.s6+
J36
or 40 + (1.96X0.75)
or 40 + 1.47 years
Illustration 3.2:lna random selection of 64 ofthe 2400 intersections in a small
city, the mean number of scooter accidents per year was 3.2 and the sample
standard deviation was 0. 8.
(l) Make an estimate ofthe standard deviation ofthe population from the
sarple standard deviation.
(2) Work out the standard error of mean for this finite population.
(3) If the desired confidence level is .9, what will limits
be the upper and lower
ofthe confidence interval for the mean number ofaccidents per intersection
per year?

Solution: The given ffirmation can be written as under:

N :2400 (This means that population is finite)
n :64
:3.2
o, :0'8
And the standard variate (z) for 90 per cent confidence is 1.645 (as per the
normal curve area table)
Now we can answer the given questions thus:
(1) The best point estimate ofthe standard deviation ofthe population is the
standard deviation ofthe sanrple itself
Hence
o:o:0.8
ps
(2) Standard error ofmean for the given finite population is as follows.

o, : oN;
100 Self-lnstuctional Material
fr*r/ ,_,
 Sampling Design
0.8
E4oo-64
- JA"1 z+oo-t
--v,-

0.8 836
: .,lA* NOTES
i rr*: 1o.l X0.97) :0.097
(3) 90 per cent confidence interval for the mean number of acciedents per
intersection per years is as follows:

-**,[9*." EI
-- lfi ! r'r-t J
:3.2 + (1.645X0.97)
: 3.2 + 0. 1 6 accidents per intersection.
when the sample size happens to be alarge one or when the population
standmd deviation is knowq one uses normal distnbution fo determining confidence
intervals for population nrean as stated above. But how to handle estimation problem
when population standard deviation is not known and the sanple size is small (ie.,
when n < 30)? In such situations, normal distribution is not the appropriate tool
and one can use I distribution for the purpose. While using / distribution, one
assumes that population is normal or approximatelynormal. There is a different
I distnbution for each ofthe possible degrees offreedom. When using r distribution
for estimating a population mean, one works out the degrees offreedom as equal
to n - l, where n means the size ofthe sample and then can look for critical value
of 'r' in the r distribution table for appropriate degrees offreedom at a given level
of significance. This can be illustrated by taking an example:
Illustration 3.3: The forernan ofABC mining conpany has estimated the average
quantityofironore extractedto be 36.8 tonnes per shift andthe sample standard
deviation to be 2.8 tonnes per shift, based upon a random selection of4 shifts.
Construct a 90 per cent confidence interval around this estimate.
Solution: As the standard deviation ofpopulation is not known and the size of
the sample is small, one can use r distribution for finding the required confidence
interval about the population mean. The given information can be written as trnder:
:36.8 tonnesper shift
o, :2.8 tonnesper shift
n:4
degrees of freedom : n- |: 4- |: 3 and the critical value of r for 90 per cent
confidence interval or at l0 per cent level ofsignificance is2.353 for 3 d.f (as per
the table o f I distribution).

Thus, 90 per cent confidence interval forpopulation mean is

Xt,$
',ln I

Self-Instructional Material l0l

Sampling Design
: 36.8 t23fi4
J+
: 36.8 t (2.53sX1.4)
NOTES : 36.8 + 3.294tonnes per shift.
Estimating population propoftion
As for the point estimate, the sample proportion (p) units that have a particular
chmacteristic is the best estimator ofthe population proportion ( p and its sampling
)
distribution, so long as the sample is sufficiently large, approximates the normal
diskibution. Thus, if one takes a random sample of 50 items and finds that l0 per
cent of these are defective, i.e., p: 0.10, one can use this sample proportion
(p : ofthe population proportion ( p :p : 0. I 0). In case
0.1 O)as best estimator
one wants to construct confidence interval to estimate a population proportion,
one should use the binomial distribution with the mean ofpopulation (p): n.p,
where n: number oftrials,p : probability ofa success in any ofthe trials and
population standard deviat br:
J npq. As the sample size increases, the binomial
distribution approaches normal distribution and this can be used for the purpose
ofestirnating a population proportion. The mean ofthe sampling distribution ofthe
proportion ofsuccesses (U/ is taken as equalto p and the standard deviation for
the proportion of successes, also known as the standard error ofproportion, is
taken as equal to ,tpq L
. But when population proportion is not known, then

one can estimate the population parameters by substituting the corresponding

sample statistics p and q in the formula for the standard error ofproportion to
obtain the estimated standard error ofthe proportion (shown below):

o:^l- W
p \n
using the above estimated standard error ofproportion, one can r,r,ork out
the confidence interval for population proportion thus:

pq
pt z'
n
where
p: sample proportion of successes;
q:l*P;
re : numberoftrials (size ofthe sample);
z : standard variate for given confidence level (as per normal curve
area table).
This formula can be explained as in illustration 3.4.
Illustration 3.4: Amarket research survey in which 64 consumers were contacted
states that 64 per cent ofall consumers ofa certain product were motivated by the
product's advertising. Find the confidence limits for the proportion ofconsumers
motivated by advertising in the population, given a confidence level equrl to 0" 95.
102 Self-lnstructional Material
Solution: The given information can be written as under: Sampling Design

n :64
p :64 per cent or 0.64
q:l*P:l-0.64:0.36 NOTES
and the standard variate (z) for 95 per cent confidences is 1.96 (as per the normal
curve area table)
Thus, 95 per cent confidence interval for the proportion of consumers
motivated by advertising in the population is :

n r z' ^l!!-
\n
: 0.64 tl .s6 ^/ (0'64X0'36)
:0.64 + (1.96X0.06)
:0.64 + 0_1t76
Thus, lower confidence limit is 52.24 per cent upper confidence limit is
75.76 per cent.
For the sake ofconvenience, one can summarise the formulae which give
confidence intervals while estimating population mean (p) and the population
proportion (p) s shown in the following table.
The 3.1 table summarizes important formulae concerning estimation:

Table 3.I Formulaicfor Estimation

In case of infinite In case of finite population

population
Estimating population mean
(p) when we know o, )- + z.--L
C'

J,
xxz.2*.W
Jn N-l V
Estimating population mean
(ir) when we do not know oo
and use o. as the best estimate X-6 + z.:L xxr.fr*,H
of o, and sample is large (i.e.,
n>30)
Estimating population mean
(p) when we do not know o,
and use o. as the best estimate
xt_t In - o
x+t.--!x l-N-;
of o, and sample is large (i.e.,
^'l ^ln \,rY-t
n?30)
Estimating the population
proportion (p) whenp is not
known but the sample is large.
otz',@
\n ,rre"rH
In case offinite population, the standard error has to be multiplied by the

finite populationmultiplier, viz., J W- n) I (N -l) .

Self-lnstructional Material 103
Sampling Design
Sample size and its determination
The most intriguing question in sampling ana\ais is: what should be the size ofthe
sample or how large or small should be r? Ifthe sample size (n) is too sma[ it may
not help achieve the objectives and ifit is too large, one would have to incur huge
NOTES
cost and resource waste. As a general rule, the sample must be ofan optirnum
size, i.e., neither be excessively large nor too small. Technically, the sample size
should be large enough to give a confidence interval of desired width. This involves
choosing the size ofthe sample thnough a logical process before sample is taken
fromteuniverse.
The size ofthe sample should be determined by a researcher, keeping in
view the following points:
o Nature of universe: The universe is either homogenous or
heterogenous in nature. Ifthe items in the universe are homogenous, a
small sample can serve the purpose. But ifthe items are heterogenous,
a larger sample would be required, Technically, this is termed as the
dispersion factor.
o Number of classes proposed: If many classes (groups and sub-
groups) are to be formed, alarge sample becomes necessary because
a snrall sarrple might not be able to accommodate a reasonable nrnber
ofitems in each class-group.
o Naturc ofstudy: Ifitemsare to be studied intensivelyand continuously,
the sample should be small. For a general survey, horvever, the size of
the sample shouldbe large. Asmall sample is considered appropriate
in technical surveys.
. Type of sampling: Sampling technique plays an important part in
determining the size ofthe sample. A small randorn sample is certain
to be much superior to a large but badly selected sample.
o Standard ofaccuracy and acceptable confidence level: Ifthe
standard ofaccuracy or the level ofprecision is to he ofa high order,
one would require relatively larger sample. For doubling the accuracy
for a fixed significance level, a fouir-fold increase in the sample size is
required.
o Availability of finance: Inpractice, size ofthe sample depends upon
the amount of money available for the study purpose. This is an
important fictorto bekept inmindwhile deterrniningthe size ofsarrple
because large sanples result in increasing the cost ofsampling estinrates.
o Other factors: Otherconsiderations include the nature ofunits, size
of the population, size of questionnaire, availability of trained
investigators, the conditions rurder which the sanple is being conductod
the time available for completion ofthe study..
A researcher has two alternative approaches for determining the size ofthe
sanple:
104 Self-lnstntctional Material
 The first approach is to specifobeforehand the precision
ofestimation that Sampling Design
is desired and then to determine the sample size necessary
to ensure it. Thesecond
approach uses Bayesian statistics to weighthe cost ofadditional
ffirmation against
the expected value from that additional information. The first
approach is capable
ofyielding a mathematical solution, and therefore, is a frequenttyusea NOTES
technique
for determining n. The limitation ofthis technique is that ii does
not analyse the
cost ofgathering information vis-d-vis the expected value ofinformation
gained.
The second approach is theoretically optimal, but is seldom used
because
ofthe difficulty involved in measuring the value offfirmation. Hence,
one needs
to mainly concentrate on the first approach.
Determining sample size through the approach based on precision
rate
and confidence level
It must be recognized at the outset that that whenever a sample study is made,
there is everypossibility of some sampling error occuring. This
is usually controlled
by selecting a sample of adequate size. A researcher will have to
specify the
precision that he.she wants in respect ofhis,&er estimates
regarding the popuiation
parameters. For instance, a researcher may like to estimate the
mean ofthe universe
within +3 ofthe true mean with 95 per cent confidence. In that case the
desired
precisionwill be +3, i.e., ifthe sample mean is { 100. the true value
ofthe mean
will t
97 andno more than { l 03. In other words, all this means
be no less than
that the acceptable
eror, e, is equal to 3. Keeping this in view, one can explain the
determination ofsample size so that specifiedprecision is ensured.
(a) Sample size when estimating a mean: The confidence interval for
the universe
mean, p, isgivenby
o
p: X + z--L
Jn
where

X: samplemean;
z: the valueofthe standard variate at a given confidence level (to be
read fromthe table giving the areas turder normal curve as
shown
in appendix) and it is r.96 for a 95 per cent confidence level.
:
n size ofthe sample
o: Standard deviation ofthe population (to be estinuted from past
expenence or on the basis ofa trial sample). Suppose, one has
o-p :4.8 for the purpose.

Ifthe difference between p and X or the acceptable error is to be kept

with
in +3 ofthe sample mean with 95 per cent confidence, then one
can express the
acceptable erto\ eas equal to

S elf- ht slruc ti o na I Materia I

Sampling Design
oD

lnr

NOTES 3 :1.96+
:ln
(1.96)'z(4.8)'z
n: :9.834 = 10.
(3)'
In general, ifone wants to estimate pr in a population with standard deviation
o_
p
with an effor no greater then e by calculating a confidence interval with
cbnfidence correspondingto z,the necessary sample size, n, equals as under:
21
z6
____;_
n=
e'
This is true ifthe population is infinite, but ifthe population is finite, the
above stated formula for determining sample size will become.

z' .N .c'o
:
' (N -l)ez + z2a2o

In case offinite population the confidence interval for pr is given by

-- x
Y+rGP @
Jn ! tn-tl
where J (ttt - n) t(N - 1) is the finite population multiplier and all other terms mean
the same thing as stated above. Ifthe precision is taken as equal to 'e', then one
has

o, N;
e: ;-f ,1i-
in J lr-t
^ ,G, N-n
e-:z-
n N-l
,'c'N 2'6'n
e,(N- 1): Jv----L-

d(N - 1) + 22o2, : z'6'rN

n

z' 'o'o'N
e'(N -l)+ z2o2o

106 Se|-lnstructional Material

 _/!s
n:- (N -l)e' + z'o', ^
Sampling Design

where
try' : size ofpopulation NOTES

n: saeofsample
e: acceptable error (the precision)
o-p : standard deviation ofpopulation
z: standardvariate at a given confidence level.
This is how one obtains the above stated formula for determining n in the
case ofinfinite population given the precision and confidence levels.

Illustration 3.5: Determine the size ofthe sample for estimate the true weight of
the cerealcontainers forthe universe withN:5000 onthe basis ofthe following
infornntion:
(1) The variance ofweight:4 ounces on the basis ofpast records.
(2) Estimate should be within 0.8 ounc€s ofthe true average weight with 99 per
cent probability.

Will there be a change in the size ofthe sample ifone assumes infinite population in
the given case? If so, explain by how much?

Solution: Inthe givenproblemwe have the following:

l/: 5000;
6
r: 2 ounces (Since the variance ofweight :4 ounces);
e: 0.8 ounces (Since the estimate should be within 0.g ounces ofthe
true average weight);
z : 2.5J (as per the table of area under normal curve for the given
confidence level of99 per cent)
Hence, the confidence interval for the universe mean, nr, is given by

x+z
# J+j
And accordinglythe sample size can be worked out rs under:

z' . N .oro
(N -l)e'z + z'o'o

Q.sT'z.(5000).(2),
(s000- 1x0.8)'z + e.s7), (2),
132098 t32098
:---:-=
3199.36+26.4196 3225.7796 = 40.95 = 4l

Self-Instrrctional Material 107

Sampling Design (a) The sample size (or n):41for the given precision and confidence
level for a finite population: Ifone takes the population to be infinite, the
sample sizewillbe calculated as under:

NOTES ,=o',, e)')

(2.s7f 26.4196
,, - e, (0.g;, 0.64
:41-28:41
In the given case the sample size remains the same even if one assumes
infinitepopulation.
In the illustration given above, the standard deviation ofthe population was
available, but in many cases the standard deviation of the population may
not be available. Since one has not yet taken the sample and is still at the
stage ofdeciding how large it should be, one cannot estimate the population
standard deviation. Inthat case where one has an idea about the range (i.e.,
the difference between the highest and lowest values) ofthe population,
one can use that to get a crude estimate ofthe standard deviation ofthe
population for getting a working idea ofthe required sample size. One way
ofgetting the estimate of standard deviation is to follow these steps:
Since 99.7 per cent ofthe area under normal curve lies within the range of
+3 standard deviations, it can be considered that these limits include almost
all ofthe distribution. Accordingly, one can assume that the given range
equals 6 standard deviations because of+3. Thus, arough estimate ofthe
population standard deviation would be:
66 : the given range

_ the given range

or
6
Ifthe range happens to be, say { 12,then
t2
o:;={2.
o

And this estimate of standard deviation, o, can be used to determine the

sample size inthe formulae givenabove.
(b) Sample size when estimating a percentage or proportion: One can
find the sample size for estimating a proportiontoo. Thereasoning put forth
would be similar to the one stated in the context ofestimating the mean.
First ofall, one has to speciSzthe precision and the confidence level and
then work out the sample size as under:
Since the confidence interval for universe proportion, is given by

pxz' p'q
n

108 Self-lnstructional Malerial

 where Sampling Design

p: sample proportion, Q
: | - p;
z : ofthe standard variate at a given confidence level and to
the value
be worked out from table showing area under Normal Curve;
n : saeof sample;
NOTES
Sincep is what one is trying to estimate, what value should one assign to it?
one method may be to take the value of p :0.5 in which case n will be the
maximum and the sample will yield at least the desired precision. This willbe the
most conservative samplestze.Theothermethodmaybeto takean initialestirnate
ofp which may either be based on personal judgment or may be the result of a
pilot study. In this case, it is suggested that a pilot study of something like 225 or
more items mayresult in a reasonable approximation ofp value. Then with the
given precision rate, the acceptable error, e, canbeexpressed as under:

-W
"'!;
or e?:zzPQ
n

or ,' . p.q
n:--:2 -
e
The formula gives the size ofthe sample in case ofinfinite population when
one has to estimate the proportion in the universe. But in case offinite population
the above formula will change to:

: z' ' p'q'N

n
;'(N -r)+ z' rq
Illustration 3.6: what should be the size ofthe sample ifa simple random sample
from apopulation of4000 items is to be drawn to estimate the per cent defective
within 2 per cent of the true value with 95. 5 per cent probability? And what would
be the size ofthe sample ifthe population is assumed to be infinite in the given
case?

Solution: The answer to the questions will be as follows:

N: 4000;
e : 0.02 (since the estimate should be within 2 per cent oftrue value);
z : 2.005 (as per table of area under normal curve for the given
confidence level of95.5 per cent ).
Since the investigator has not been given thep value being the proportion of
defectives in the universe, he/she has to assume ittobe p:0.02 (this maybe on
the basis ofone's experience or on the basis ofpast data or may be the result of a
pilot study).

Self-Instructional Material 109

Sampling Design The size ofthe sample can now be determined using aU this information for
the givenquestion as follows:

z' ' p'q'N

NOTES e'1N -t1+ z' . p.q

(2.00s)'z (0.02xI
- 0.02x4000)
(0.02)'z (4000 - 1) + (2.005)'z (0.02x1 - 0.02)

31s.1699 315'1699
-
1.5996 + 0.0788 1.6784
=187.78= 188

And ifthepopulation is infinite, thenthe sample size willbe as under:

z' ' p'q

: ----;-
n
e'
(2.00s)'z .(0.02x1-0.02)
_
Q.OD,
0.0788
_ ________: = l g6.gg =lg7
0.0004
Illustration Letus assume that a certain hotel rnanagement is interested in
3.72
determining the percentage ofthe hotel's guests who stay for more than 3 days.
The reservation manager wants to be 95 per cent confident that the percentage
has been estimated to be within +3 per cent of the value. what will be the
conservative sample size needed to tackle this problem?

Solution: Ifthe population is infinite:

e : 0.03 (since the estimate should be within 3 per cent of the true
vahre);
z : 1.96 (as per table of area under normal curve for the given
confidence level of95 per cent)
As one wants the most conservative sample size one can take the value of
p:0.5 ndq:0.5. Using allthis information, one candeterminethe sample size
for the givenproblem as under:

n: z'pq
t
(1.e6)'? .(0.sx1- 0.5) 0.e604
= 1067 .11 = 1067
(0.03)'? 0.0009

Then, the most conservative sample size needed for the problem is : 1067.
Determination of sample size through the approach based on Bayesian
Statistics:
110 Self-lnstructional Material
 Sampling Design
Another approach to determine zl is to use the Bayesian statistics known as
the Bayesian approach. The procedure for finding the optimalvalue ofz or the
size ofsample under this approachis as under:
(i) Find the expected value ofthe sample information (EVSD for every
NOTES
possible n;
(i| Workout reasonably the approximate cost oftaking a sample ofevery
possible n;
(iif Conrpare the EVSI and the cost ofthe sample for every possible r. In
other words, work out the expected net gain @NG) for every possible
n as stated below:
For a given sample size (n): (EVSI) - (Cost of sample) : (ENG)
(iv) From (iii) above, the optimal sample size, that is the value ofn which
maximizes the difference between the EVSI and the cost ofthe satryle,
canbe determined.
One disadvantage ofthis approach is that the computation of EVSI for
everypossible n andthen comparing the same with the respective cost is often
a very cumbersome task and is generally feasible with the help ofcomputers only.
Therefore, although theoretically optimal this approach is rarelyused in practice.

CrmcxYouR PRocREss

8. What is the concept ofstandard error?

9. What is the central limit theorem?
10. What does the Bayesian statistic suggest?

3.4 SUMMARY

o Universe or population in terms of statistics includes all items in the

researcher's field of inquiry. A sample is the small number of elements
selected to represent the entire populatioq according to some rule or plan.
A measure based on a sample is known as a statistic.
o To measure the characteristics ofthe miverse, a small sanple that is assumed
to nearlyrepresent the entire population is studied. And then the results are
projected over the entire population to get the approximate overall picture.
. oftwo broad types-probability sampling
Sampling methods are basically
and non-probability sampling. Again each broad type is divided into
subtypes.
o It isnecessaryto learndifferent sampling elrors andhowto minimize those
errors. There are two types oferrors that can affect the inferences made in
findings--sptemic bias and sampling elrors.
Sef-lnstructional Material lll
 Sampling Design o Slatemic bias can occur due to inappropriate sampling, defective measuring
device, non-responding, indeterminancy and natural bias in the reporting o1
the data. Sampling eruors arise from inaccurate sampling and they generally
happen to be random variations (when sampling is random) in thi samp6
NOTES estimates, around the true population values.
o To get the right inferences it is necessaryto keep the sampling and systemic
errors at the minimum possible limits.
o Probability sampling methods include simple random sampling, stratified
random sampling, cluster sampling, quota sampring, etc. Non-probability
sampling methods include judgmental sampling, accidental sa- pling,
systematic sampling, snow-ball sampling, etc.
o A researcher is basically interested in three types ofdistributions-sample
distribution, population distribution and sanpling distnbution.
o To get the sample distribution the observed data are arranged in a frequency
distribution, e.g., fertilityrate. Then, forexample, ifthe fertilityrates ofall
the 30,000people ofthepopulationare obtained and arrangedin a frequency
distribution, it is known as the population distribution. Since the forms rnd
parameters are not ordinarilyknowq an estimate ofthese two characteristics
ofthepopulation is made fromthe sample distribution. So, ifthe sample
distribution is normal, one can assume that the population distribution is
also normal.
o In statistics, a sampling distribution or finite-sample distribution (which is
different from sample distribution) is the distribution ofa given statistic based
on a random sample of size n. It may be considered as the distribution of
the statistic for all possible samples from the same population ofa given
size. The sanpling distribution depends on the underlying distnbutionofthe
population, the statistic being considered and the sample size used. The
sampling distribution is frequently opposed to the asymptotic distribution,
which corresponds to the limit case n -> @.
. Sampling distnbutions: Important sampling distnbutions that are commonly
used are: (i) sampling distribution ofmean, (ii) sampling distribution
of
population, (iii) Students t distribution, F distribution and chi-square
distnbution.
o while conducting a research, the researcher should airn at
getting the most
accurate results possible, subject to the constraints ofmoney, time, facilities
andeffort.

3.5 KEY TERMS

o systemic bias: Abias ofa measurement system or estimate method.
a F distribution: A probability distribution of the ratio of two variables.
a chi-square (26'?) Aprobability distribution used in inferential statistics.
l12 Self-lnstntctional Material
 student's I test: This test is used when two conditions are fulfilled: the Sampling Design
sample size is 30 or less and the population variance is not
known.
Sandler's I test: Atest that serves the same purpose as is accomplished
by I test relating to a paired data.
NOTES
3.6 ANSWERS TO 'CHECK YOUR PROGRESS'

l. when all items ofa universe are enumerated, it is called a census inquiry.
2. A simple random sample is also known as an unrestricted random sample.
Here every individual of a population has an equal chance of
being
included in the sample.
3. Probability sampling is preferred because it ensures that every element
of
the population has an equal chance ofbeing included.

4. Random sampling means all the elements of a population have

an equal
chance ofbeing included in the sample. The methods ofdrawing
random
sample are: O Fishbowl drawmethod; (ii) Using table ofrandomnumbers;
and (iii) Computer determined randomness.

5. Quota sampling is a non-probability sanpling method. Here different strata

are identified and from each stratum the number of individuals
needed is
selected forthe study.
6. In behavioural research people tend to give 'right, rurswers, as opposed to
true'answers.'
7. cluster sampling is a form of probability sampling. Generally
large
geographicalregions,territories, conrrnrnity, neighborhood,
etc. are identified
on a map by a number and then from these, a sample is
drawn for the study.
8. Standard error hess in testing whether the diference between
the observed
and the expected frequencies arise due to chance.

9. when sampling is not from a normal population and the size ofthe z,
the
shape ofthe distribution will depend on the parent population.
But as r gets
larger, the sampling distribution would resemble a normal
distribution,
irrespective ofthe shape ofthe population distribution.
l0' The Bayesian statistic suggests weighing the cost ofadditional
information
obtained against the expectd value ofthe additional information
The second
approach is difficult to determine, so the fust is used often.

3.7 QUESTTONS AND EXERCTSES

Short-Answer Questions
1. write two points to be kept in mind while designing a sample.
2. Define bias and sampling errors.
Self-lnstmctional Material ll3
Sampling Design 3. What are probability sampling methods?
4. What is sampling distribution ofmean?

5. What is sampling distribution ofpopulation?

NOTES Long-Answer Questions
1. Explain the different types of sampling designs and their advantages and
disadvantages.
2. Explain the three types ofstatistical distributions in detail.

3. Explain student's rdistnbution.

4. Write a note on Fdistribution.
5. Explain Chi-squde distribution.

3.8 FURTHER READING

Cochran, W.G. 1963. SamplingTbchniques. NewYork: JohnWileyand Sons

Kerlinger, F.N. 1973. Foundations of Behavioral Researcft. NewYork: Holt
Reinhart and Winston, Inc.
Kothari, C.R. 1995. Research Methodology: Methods and Tbchniqzres. New
Delhi: Wishwa Prakashan.
Singh, A.K. 2008. Tests, Measwements and Research Methods in Behavioural
Sciences.New Delhi: Bharati Bhawan.
Siegel, S. 1956. Nonparametric Stotisticsfor the Behavioral Sciences. New
York McGraw-Hill Publishing Co., Inc.

114 Self-lnstructional Material

 4 RESEARCH
Research Design
UNIT DESIGN
Strucfure
NOTES
4.0 Introduction
4.1 Unit Objectives
4.2 Research Design
4.3 Experimental Design
4.3.1 Important Experirnental Designs
4.3.2 Quasi-Experimental Designs
4.4 Qualitative Methods
4.4.1 Definitions of eualitative Research
4.4.2 Distinctions between eualitative and euantitative Research
4.4.3 Research Perspective
4.4.4 Texts and Realities in eualitative Research
4.4.5 Theories in Qualitative Research
4.4.6 Sarnpling Strategies in eualitative Research
4.4.7 Types of Interviews for eualitative Research
4.4.8 Focus Groups
4.4.9 Msual Data
4.4.10 Ethnography
44.11 CodingandCategorizing
4.4.12 Content Analysis
44.13 Documentation ofData in eualitative Research
4.4.14 Presenting the Findings/Results Following the Qualitative Research
4.4.15 Computers in Qualitative Research
4.5 Summary
4.6 Key Terms
4.7 Answers to 'Check Your Progress'
4.8 Questions and Exercises
4.9 Further Reading

4.0 INTRODUCTION

In this unit, you will learn the meaning and the purpose of research design. A
research design is a plan and a systematic procedure for collecting the data and
performing analysis on that data for the purpose ofresearch. In other words, a
research design is a conceptual framework for conducting research. It is a blueprint
for collecting, measuring and atnlyzngdata. Research designs tell us what, where,
whenandhow anyinquiryisto be made.
The unit explains the features and important concepts related to research
design. It throws light on different bapes ofresearch designs. It explains the basic
principles and different types of experimental designs. It also discusses quasi-
experimentaldesign.
The later part ofthe unit teaches you the principles and different types of
qualitative research. It will also make you familiar with different types ofinte*i"rr,

Self-Instructional Material ll5

Research Design group discussions and observations. At the end ofthe unit, you will also leam data
documentation and analysis.

NOTES
4.I UNIT OBJECTIVES
Aft er going ttrough this unit, you will be able to :

o Get familiar withthe concept and tSpes ofresearch designs

o Understand sampling designand its importance
o Learn the basic principles and different types ofexperimental design
o Know the concept and tlpes ofqualitative research
o Examine content ana$sis in qualitative research

4.2 RESEARCH DESIGN

A research design is a conceptual framework for conducting research. It is a

blueprint for collecting, measuring and analynng data. Research designs deal with
the What, Where, When and How ofan inquiry. Several questions in the following
list have to be answered before starting a research work.
. What is the study and who is doing the research?
o Whyis the studybeing conduced?
o Where willit be done?
o What tlpe ofdata is to be collected?
o What is thetimeframe forthe study?

. o What techniqueswillbe used fordata collection?

o Honwillit be analyzed?
o What kind ofreport willbe prepared?
o What willbe the costs involved?
. Who are the personnel for doing the study?
Keeping these in mind, the design can be sub-divided into specific areas for clarity:
o The sampling design: Methods ofselecting the items for observation for
thestudy
o The observational design: The conditions under which the observations
are to be carried out.
o The statistical design: The data gathering and analysis methods.
o The operational design features: The entire procedure for carrying out
the study.

116 Self-Imtructional Material

Important features of a design Research Design

The important features ofa design are that it is:

o A plan that identifies the resources and the type ofinformation needed
o The strategyforgathering data NOTES
o An estimate oftime, costs, etc.
So, the research design needs to have a clear research problem procedure
for data collection, the population to be studied and the type ofdata analysis that
has to be carried out.

Need for a research design

For smooth conduct ofthe study, for efficient data gathering and analysis and for
economyand effectiveness, research designs must be plannedwell in advance
and with great care. Good designs help to obtain reriable results.

Features of good design

o It should seek to minimize bias and maximize reliability ofthe data obtained
o It should give the least possible experimental error
o It should be as objective as possible
No single design should be used for all ffis
ofresearch problems. Reporting
on the purpose ofthe study, the type ofdata needed and other considerations
determine the designto be chosen.

Important concepts relating to research design

1. Dependent and independent variables: Avariable is anyunit that can
have different quantitative values, e.g., height, loudness etc.
Qualitative units
are attributes, e.g., honest, extroverted, shy, etc. Largely quantitative
variables are continuous, e.g., age is a continuous variable while ,students'
is a non-continuous variable.
(a) An antecedent variable is an independent variable.
(b) A consequent variable is a dependent variable.
(c) Height is a variable dependent on age where age is an independent
variable
(d) Height is sex related, so height is a dependent
(e) Variable, age and sex are independent variables.
(f) Behaviour changes as a function ofthe manipulation is an independent
variable.
(g) The independent variable (IV) is the one that is manipulated. It is
under the control ofthe experimenter, generally. This variable is also
called the experinrental variable. The effect ofthe e4perimental variable
is reflected on the dependent variable, e.g., knowledge of results
improves learning. Here providing knowledge of results are the

Self-Instructional Material ll7

Research Design independent variables (DVs) and the improvement in learning is the
dependent variable.
2. Extraneous variables: Variables uffelated to the study but having an
influence on the dependent variable are called extraneous variables.
NOTES
Examination results studied to be a function ofthe methods of studying.
Results are DVs and the methods of study are IVs. However, intelligence
also plays apart in the results. This becomes an extraneous variable, and
affects the outcome. Such an influence is known as the experimental error.
3 . Control: This implies any atternpt to minimize the inftuence ofthe extraneot$
factor(s) or variable(s). The attempt is to keep the experimental conditions,
wellcontrolled.
4. Confounding or compounding influence: When the extemal frctors cannot
be controlled, they are thought to compound the outcome(s).

5. Research hypothesis: When a hypothesised relationship is to be tested

scientifically, it is called fupothesis testing. Ahypothesis is a hunch. It should
have one IV and one DV, as part of the design.
6. Experimental and non-experimental hypotheses testing research:
When a hypothesis is to be tested, it is known as a hypothesis testing
research. Here the experimentalvariable is manipulated. Whenthe IV is
not manipulated, it is called as non-experimentalhypothesis testing research.'
For example, how age affects memory. Here people ofdifferent ages are
tested on a memory task. The memory is determined by calculating a
coefficient ofcorrelation between the sets ofobtained scores. This is a non-
experimental hypothesis testing research.
7. Experimental and control groups: The control group is exposed to
the regular conditions. The experimental group is exposed to the
experimental variable. The experimental group receives the experimental
variable. This enables comparison in terms ofthe outcome on the dependent
variable.
8. Treatments: The conditions under which the two groups are studied is the
treatment procedure.
9. Eryeriment: The procedure for testing a hlpothesis or atterrpts to establish
the veracity ofa relationship is known as an experiment.
10. Experimental unit(s): The pre-determined block or conditions where
different treatments are applied is called an experimental unit.

Different types research designs

Designs canbe categorized as:
(a) Exploratorystudydesign
(b) Descriptive studies design

118 SeA-lnstructioilal Materiql

 (c) Diagnostic studydesign Research Design

(d) Hlpothesis-testing study design

Exploratory study design: The purpose here is to do a preliminary study
to be able to formulate a research design later. Such a design needs to be flexible
NOTES
to be altered, depending on what the explorations yield. These designs are based
on survey ofthe literature, survey ofthe experiences encountered, and anallsis of
insights or intuitions.

Research design for descriptive and diagnostic research: These

designs must be tight and well planned.
Steps in this type ofdesign:
o State objective(s) and design methods ofdata collection
o Select samples, suitably colle ct data, analyzeand report results.
o Questionnaires, interviews, case-studies and observations are used
extensively, here.

Tqble 4.1 Summary of Research Designs

Type of Study
Research Design

Exploratory or Formulative Descriptive/Diagrrostic

Flexible design (design Right design (design must

must provide opportunity make enough provision
Overall Design for considering different for protection against bias
aspects of the problem) and must maximize
reliability)

(i) Sampling design Non-probability sampling Probability sampling design

design (purpose or (random sampling)
judgment sampling)

(ii) Statistical design No pre-planned design Pre-planned design for

for analysis analysis

(iii) Observational Unstructured instrument Structured or well thought

design for collection ofdata out instruments for
collection of data

(iv) Operational No fixed decisions about Advanced decisions about

Design the operational procedures operational procedures

Source : Kothari, C.R., Research Methodologt Methods and Techniques, 1995.

Research design in hypothesis testing: Here the researcher tests the

causal relationship between variables. Studies have to help in reducing bias and
increase reliability.

Self-lnstructional Marerial ll9

Research Design Basic principles of experimental design:
o The principle ofreplication
o The principle of randomization
NOTES o The principle oflocal control
The principle of replication: It means that any experiment should be repeatable.
No single performance ofa study is sufficient for results to be reliable. Also, the
precision ofa study is enhanced with replication.
The principle of randomization: This method is to protect against
extraneous factors creeping into and contaminating the outcome(s). The role of
chance factors is sought to be neutralized through randomization.

The principle of local control: The design is so structured that the

extraneous factor influences are controlled. Here, the field of study is divided into
homogeneous parts equal to the number oftreatments. The treatments are then
randomly assigned to these blocks. A block is a division of the field into
homogeneous parts. The blocks are the levels at which the extraneous factor is
fixed. This can lead to the elimination ofthe variability due to the extraneous factor(s)
that cause the experimental error.
There are several research designs possible. The researcher has to choose
a design keeping in mind the objective ofthe study, the tlpe ofdata needed, the
sanplingmethods, the desired levelofaccuracyand other constraints underwhich
the studyisdone.

Developing a research plan

The research plan is the writhen format ofthe research that is to be undertaken.
Thisisuseful:
o For organizing thoughts, ideas and concepts, clearly and well
o For develop a checklist of all materials needed for research.
o As a preparatory document that is available for scrutinli
The plan must consist of:

0 The research objective

(ir) The problemto be investigated
(ii) The majorvariables and issues to be studied/measured
(iv) The methods /techniques to be used
(v) The proceduresto be followed
(vi) The sample: fromwhere to be drawn, the size, etc.
(vii) Statistical and othertools to be used
(vii) Budget and time constraints involved.

120 Self-lnstructional Material

 Research Design

CrmcrYouR Pnocnrss
1. What isadesign?
2. What is a variable? NOTES
3. Name are the different research designs?
4. What ismeant byrandomization?

4.3 EXPERIMENTAL DESIGN

4.3.1 Important Experimental Designs

E4perimental designs are the structures ofan experiment, which are oftrvo tlpes-
informal and formal. Informal designs have less sophistication and less control
while the formal designs offer more control and lend themselves to the use of
precise statistical procedures for analysis.

The informal experimental designs:

(i) Before-and-after without control design
(if After-only, withcontrol desrgn
(iif Before-and-after with control design
The formal experimental designs :

(i) Completelyrandomized design (C-R design)

(if Randomized Block design (R-B design)
(ii) Latin- square design (L-S design)
(iv) Factorial designs
4.3.1.1 The informal experimental designs
(i) Before-and-after without control design: Here a single test group or
area is selected and the dependentvariable is measuredbeforethetreatment
or intervention is introduced. Then the variable or the intervention
is introduced and again the dependent variable is measured. The effect of
the treatment would be equal to the level of a given phenomenon after the
treatment minus the levelofthe phenomenonbefore the treatment.
The main dfficulty with such an experiment is that with the passage oftime,
some extaneous variations could enterinto the heatmenteffect, e.g., fatigue,
boredor4 familiarity, etc.
(ii) After-only with contrcl design: In this design, two conditions are involved
(experinrental condition and control condition). The treatrnent or intervention
is introduced into the experimental condition only. The dependent variable
is then measured in both the conditions at the same time. The treatment

Self-Instructional Material l2l

Research Design effect is obtained by subtracting the value ofthe dependent variable in the
control condition from the value in the experimental condition.

Experimental conditi on Treatment introduced V'alue of the dependent variable (r)

NOTES after treatment
Control condition Treatment withheld Value of the dependent variable
without treatment (y)
Treatment effect: (x) * (y)

S o urce : Kothari, C.R., Re s earc: h Meth o d o logy Met h ods and Te c hn iqu e s, 199 5.

The basic assumption ofthis design is that the groups are identical with
respect to their behaviour towards the phenomenon under consideration. If
this is not tnre, there are chances of extraneous factors entering into the
treatment Problems with respect to time lapses do not enter into this design.
This design is superior to Before-and-After without control design.
(iii) before-and-after with control design: In this design, two groups are
selected and the dependent variable is measured in both, for an identical
length oftime prior to introducing the independent variable or treatment.
Then the independent variable or the treatment is introduced only in the
experimental condition (as shown in the table below). After an identical
tinre period lapses, the dependent variable is measured in boththe conditions.
The treatment effect is determined by subtracting the change in the dependent
variable inthe control condition frornthe change in the dependent variable,
in the experimental condition.

Table 4,2 Befae-And-Afterwith Contrul Desigt

Time Period I Time Period lI
Test area: Level clfphenomenon Level ofphenomenon
before treatment (Ji.1 introduced after treatment (l)
Control area: Level ofphenomenon Level ofphenomenon
Without treatment without treatment
(A) @
Treatment effect : U- n - Q - A)
This design is superior to the earlier two designs in that it overcomes
extraneous influences arising from time lapses and also from the non-comparability
ofthe experimental and control conditions. However, when suitable control groups
are not available, one ofthe other two earlier designs could be considered.

4.3.1.2 The formal experimental designs

(i) Completely randomized design (CR design): This design rests on the
principle of randomization. The subjects are assigned randomly to the
experimental conditions.
Two-group simple randomized design: In this design, the population is
identified and then a sample is chosen randomly. Then the subjects are
122 Self-lnstructional Material
 again randomly assigned to the experimental and control group
conditions. Research Design
This ensures the principle of randomization, so both the groups are
representative ofthe population chosen for the study.

NOTES

I *ffi].*i13"13jr-{'#r- l-, i::1J#r,

Fig.4.1 Two-group Simple Design (in Diagram

^*OrTT,l,Experimental

Since the sample is drawn randomly from the population and then again
randomlyassigned to the two conditions (experimentaland control) and
then each ofthese conditions receive different treatments (.r andy), the
independent variables; the conchrsions drawn from the samples are applicable
to the population. Such a design has the merit ofrandomizing the sample.
one could test each goup before and after the treatment afteiensuringlhe
equivalence ofthe two groups.
Individual differences that exist in the two conditions and the experimenter
influences (e.g., theteacher ortrainer differences inapplying thi methods)
can be further controlled by the random replication deiign. such diftrences
get minimized with such a design.

(ii) Randomized block design (RB design): This is an improvement over

the completelyrandomized design. Here, the subjects are first divided into
two groups known as blocks (married men, single meq high or low incorne,
etc.). So, the blocks are homogeneous with respect to the chosen variable.
The number ofsubjects in a given brock would be equal to the number of
treatments. one subject in each block would be randomly assigned to each
treatment. The blocks are the levels at which the extraneous frctor is fixed.
This enables one to know its contribution to the total variation. The main
feature ofthis design is that each treatment appears the same number of
times in each block. The RB design is analysed by a two-wayANovA
(TheANalysisOfVariance) ]

(iii) Latin-square design (L-S design): A Latin square is a multi-faceted

experimental design. The Latin square can be used in many different
situations. It is an elperinrental design that consists ofallocation oftreatnpnts
to eryerimental tlrits. The set ofexperimental urits and the set oftreafinents
may be structured in some way.
Below is a Latin square oforder 5. In a number ofdifferent experinrents this
could be the basic structure for the design. The following three examples
will help the student understand the concept oflatin squares better.
Sef-lnstmctional Material 123
Research Design First example
Suppose we want to test five drugsA, B, C, D and E for their efficacy in alleviating
the synrptoms ofa chronic disease. Five patients arc available for a trial and each
willbe available for five weeks. Testing a single drug requires aweek. Thus an
NOTES
experimental unit is a'patient week.'
The stucture ofthe experimentalurits is a rectangular grid (which happens
to be square as in Fig. 4.2).There is no structure for the set oftreatments. We can
use the Latin square to allocate treatments. Ifthe rows ofthe square represent
patients and the columns are weeks, then for example, the second patient in the
third week ofthe trial, willbe given drug D. Now eachpatient receives all five
drugs and thus in each week, all the five drugs are tested.

A B C D E
B A D E C
C E A B D
D C E A B
E D B C A

Fig, 4,2 Experimental units

Second example
The second exanrple is very similar to the first. Suppose we are testing five varieties
ofpesticides on a square orchard with 25 apple trees. There may be differences
between the rows, and differences between the columns, but we assume that
rows and columns have the same status (e.g., the orchard is not on a hillside).
Each plot occurs in one row and one column, just as each experimental unit in the
preceding example occurs in one patient and one week. In a sense, the rows are
'equivalent' to the colurnns, but the patients are not equivalent to weeks in the
same way as in the earlier example, because of the influence of the drug. The
structure ontheplots is a Hamming scheme inthis case.
Again the Latin square gives an allocation ofvarieties to plots in the same
way as it does in the earlier case.
Third example
Suppose we have to test five different cake recipes and five baking ternperatures.
We canperform experiments on five days, and five timeslots are available each
day. The 25 experimental units are divided into five blocks of five, while the
treatments have a 5x5 factorial structure.
Again one can assign treatments to units using the Latin square. Let us
number the cells ofthe square from I to 25 as shown in Fig. 4.3.

I 2 J 4 5
6 7 8 9 l0
ll t2 l3 t4 l5
t6 t7 l8 t9 20
2t 22 23 24 25

Fig.4.3 Latin Square

124 Self-lnstructional Material
 we assume that the rows ofthe table represent days, so that units 1 to 5 are Research Design

the timeslots on the first day. Now we use the five columns ofthe square to assign
the five temperatures, and the letter to assign the recipes l, B, C, D, E.
(iv) Factorial design: This is used when to study the effects byvarying more
NOTES
than one factor. This is especiallyuseful in social science research where a
number offactors do influence a particularphenomenon. Factorial designs
are of two types: l. Simple factorial designs, and2. complex factored
designs.

A simple factorialdesign is one where the effects ofvarying two factors on

the dependent variable are studied. Ifmore thantwo factors are involved,
the complex factorial design is used. A surple factorial design is also known
as a two-factor frctorial design while a complex factorial design is called a
multifactor fr ctorial design.

Experimental variable

Confol variable Treatment A Treatment B

Level I Cell I Cell3

Level II Cell2 Cell4

Fig 4.4; 2 x 2 Simple Factorial Design

Here there are two treatments or experimental conditions and two control
conditions. This sample is divided into four cells. Each ofthese has one
fteatment condition Subjects are assigned at randorn The means for different
rows and columns can be obtained. The column means represent the main
effects oftreatments. The row means are the main effects for level without
regard to treatment. So, the main effects oftreatment as wellas levels can
be studied by this design. Further, the interaction between treatments and
levels canalso be studied. This enables one to seewhetherthe treatments
and levels are independent ofeach other or not.
complex-factorial design: A complex-factorial design has more than two
factors at a time. It is a design with three or more than three independent
variables simultaneously.Inthe following design, there are (3 l) factors.
-
Experimental variable has two treatments and two control variables each
with two levels. Such a design is a2x2xz complex frctorial design. There
are atotalofeight cells. The design is as givenbelow:

S e lf- In s tru c ti o na I Mat er ia I

Research Design Table 4.3 2 x 2 x 2 Complex Factorial Design

Experimerrtal Variable
TreatmentA Treatment B
Control Control Control Control
NOTES Variable Variable
Variable Variable
2 2 2 2
Level I Level ll Level I Level II
Cell3 Cell5 Cell
control J-- Level I Cell I T

Variable I
l- Level II
I
Cell2 Cell4 Cell6 Cell8

Source: Kothari, C.R., Research Methodo I ogy Methods and Techniques, 1995.

Fromthis design, it ispossibleto determinethemaineffectsofthesevariables:

one experimental and two control variables. The interaction between these
pairs ofvariables can also be determined. Such an interaction is called a
first order interaction. And when the interaction between variables, taken in
triplets, it is called a second order interaction.
Experimental variable with control I (or EVxCV I );
Experimental variable with control variable2 (or EVxCV2);
Control variable 1 with control variable 2 (or CV 1 xCV2):
There will be one second order interaction in this design. It involves all the
tlree variables, i.e., EVxCV 1 xCV2).
To determine the main effect for the experimental variable, one needs to
compare the cornbined mean data in cells 1,2, 3 and 4. For treatment A and the
combined mean data in cells 5,6,7 and 8 for treatment B. In this way, the main
effect ofthe experimental variable is obtained without the influence ofcontrol
variables I and2. Similarly, the maineffect ofcontrolvariable 1, independent of
the experimental variable, and control variable 2 can be obtained ifone conpares
the combined mean data in cells 1,,3,,5 and 7 with the combined mean data in
cells2,,4,,6 and 8 in the2x2x2 design. Similarly, one can determine the main
effect for control variable 2 also independent ofexperimental variable and control
variable l, ifthe combined mean data of cells 1,2,5 and 6 are combined with
mean data ofcells 3, 4,7 and 8. To obtain the first order interaction for EVxCVI
in the above research design, the researcher can ignore control variable 2 and
generate a 2x2 design from the two 2x2x2 design by choosing to combined data
fromthe relevant cells as shownbelow:

126 Self-lnstructional Malerial

 Table 4.4 Expemental Variables Research Design

Experimental Variables

Treatment A 'l'reatrnent B
NOTES
Level I Clells 1,3 Cells 5. 7
Control
Variable I
Le.r,el ll Cells 2. 4 Cells 6, 8

so urc e : Kothari, c.R.. Re s earc h Me th o do I ogy Me t ho ds a ncr rb c hn i ques, 1995.

The same can be done for other fost order interactions. The complex frctorial
design need not be a 2x2x2 design only. Any number of combinations of
experimental and control variables, which are independent, can be used for
generating data and analysis. The more the rariables, the more complex the design
Factorial designs have two crucial advantages: l. Equivalence accuracy:
One can determine the rmain effects oftwo or more factors in one single experinrent.
It is economicalin terms ofeffort, 2. This design also permits various comparisons
of interest to the researcher, i.e., certain effects that cannot be as ascertiirr.A uy
treating one factor at a time. The detennination ofinteraction effects is also possible
with the use ofthis design. The researcher must decide in advance what the most
appropriate design is for a given study.

4.3.1.3 Other important types of experimental designs

The designofan experiment is a blueprint oftheprocedures that are to be followed
for testing a hypothesis. Ifthe experimenter decides that there would be only one
group of subjects who will be tested under the different values or conditions or
treatment ofthe independent variable; the ensuing design would be known as the
'within groups design' or 'repeated treatment design'. If the decision is to use
separate groups for each value of the independent variable; they are also know as
between subjects orwithin subjects designs.

Experimcntal Design

Between-groups Within-groups
design design

Randomized-groups N.{atched-groups pactorial uplete lncomplete

design clesign design

Two-randomized-groups Morc-tnan-two-ranciomized-groups
design desigl

Fig 4.5: Experimental Desigtt

Source: Singh, A.K., Iesrs, Measurement and Research Methods in Behavioural Sciences,
2008.

Based on the number ofgroups, the two designs are: 1. Between groups and 2.
within groups. ln psychological and educational studies the 'between groups design,
Self-Insrructional Material 127
Research Design is used often. 'Between group designs' are divided into the following three types:
a. Randomized groups design
b. Matched-goup design
NOTES c. Factoreddesign
Between groups designs
Randomized groups design: It is one in which subjects are assigned randomly
to different groups meant for the different conditions or values ofthe independent
variable. The assumption here is that the random assignment makes these groups
statistically equivalent. This means the variations in the dependent variable become
easily identifiable. When the subjects are randomly assigned to onlytwo groups;
the design is called as 'two-randomized groups design. And when the subjects are
assigned to more than two groups, it is called a multi-group desrgn'.
Matched-groups design: It is another between groups design in which
the subjects are matched on mean, standard deviation, pairs, etc.
Factorial design: In this design two or more independent variables are
studied in various possible combinations. Here their independent and interactive
effects on the dependent variable can also be studied.
There are two primary ways through which unbiased groups or random
groups subjects can be formed: 1. Captive assignment, and2. Sequential
assignmart.
Captive assignment: Here allthe subjects are individuallyknown and for
the duration ofthe experiment they are made captive, so that they are randomly
assigned to the different conditions or groups and so the term captive assignment.
The randomprocedure is followed. The onlypre-requisite is that the 'M ofthe
groups, be equal.
Sequential assignment: The technique of sequential assignment is one
where the experimenter does not know the subjects in advance. As the experiment
progresses; the experimenter may use three objects on day one, six on day two
and so on. So the experiment follows a pre-arranged schedule or sequence.
The sequential assignment ofpre-arranged schedule or sequence or simply
the sequentialassignment of subjects canbe done by complete randomization or
block randomization. Any biases that could arise in the sequence can be overcome
byblockrandomization. The concept ofblockrandomization is thatthe independent
variable occurs in each successive block oftrials once, but the order ofconditions
within a block is random and different from every other condition. All conditions
must occur equal number of times for the block randomization to be effective.
Social sciences researches have to establish which ofthe several independent
variables influences the dependent variable most, and also the type ofrelationship
that exists between the most influentialof independent variables andthe dependent
variable. At'wo-group randomised design satisfies the first nee{ but not the second
one. Here two independent variables are involved so it is difficult to ascertain the
f28 Self-lnstructional Material
 most significant independent variable on the dependent variable. It is irnportant to Research Design
studymore thantwo values ofthe independent variable forthispurpose. This is
the 'more thantwo randomized group dergn'.
Two-group randomized design: The subjects are assigned randomly to
the two groups. The independent and dependent variables have to be defined NOTES
clearly. The two values ofthe independent variable have to be identified. The
values are conditions ortreatments ofthe e4periment. The objective ofthe studyis
to see whether the two conditions affect the dependent variable.
The populationhas to be specified and the sample drawn. The subjects are
then randomly assigned to the two groups. (Method ofrandom selection to be
followed)
Here, two equalNgroups are formed. Acoin toss could decide which the
experimental group is and which the control group is. The experimental group
receives the experimental treatment, while the control group does not receive it.
After the experiment, the scores obtained by the subjects in both the groups on
the dependent variable are recorded. Usually the r test or the non-parametric
Mann whitney utest is used in a two-group randomized design. Ifthe two groups
differ, thenthe independent variable is seento affect the dependent variable. St,
the independent variable manipulation is seen to influence the resultant dependent
variable.
central to the two-group randomized design is the random assignment of
subjects into the groups, so that no systematic relationship between the
characteristics ofthe subjects and a particular group to which they are randomly
assigned emerges. To achieve this, (underwood, 1966) suggests two ways in
which random groups can be formed.
i Captive assignment
ii Sequential assignment
More-thon-two-randomized-groups design: This is also called a multi-
groups design Here, there are three ormore values or conditions ofthe independent
variable. So, three or more groups of subjects participate. All the subjects are
randomly assigned to the three ormore groups. The process ofcaptive assignrnent
or sequential assignment using the random procedure can be used here. For exanple,
for hunger drive levels on responses, three groups of experimental variables with
three conditions ofdeprivatiorr-mild, moderate and extreme deprivation offood
can be chosen. Ifthe two-groups design was chosen and the conditions were mild
and moderate deprivation and the number of correct responses (bar press) was
used as the dependent variable, perhaps the two conditions mild and moderate
would not yield a significant difference. So, the use ofthe third group (extreme
deprivation) would have a clear effect on the dependent variable. This rezult would
have been missed in the two-group experiment.

Multi-groups design helps in determining the relationships between the

independent and the dependent variables that cannot be established by the two-
group design. l

Self-Instructional Material 129

Research Design Matched groups design.' This is also known as the randomized block
desrgn. It can have two or more groups. All the subjects are given a common pre-
test on a task (matching variable) and then are formed into groups on the basis of
theirperformance. This ensures equivalence ofgroups. Thenthe different values
NOTES ofthe independent variable are introduced to each group. Ifthese groups have
had equivalentmeans onthe dependentvariablepriorto theexperinrentaltreatment
isgrer,, then ifthe significant difference occurs, after the given, and then if significant
difference occurs, after the introduction ofthe treatment, the resulting differences
may be confidently be attributed to the experimental treatment.
This design assumes that the subjects in a group are homogeneous. They
form block. So it is also known as a randomized block design.
a

Two issues connected to the matching task:

(a) Selection ofthe matching variables
(b) Waysofmatching
The subjects are pre-tested on a matching variable in this design. The
rnatching variable is chosen on the basis ofits abilityto yield a high conelation with
the experimentaltask, or the dependent variable. To obtain this, the dependent
variable itselfis used as the matching variable, for exanple, to studyknowledge of
results onmemory the subjects are all given a memorytask and a set ofscores are
obtained---one with knowledge ofresults and the other without it. After the pre-
test, they are paired off The method for pairing into the two groups is through
randomization. Then the experiment is carried out, as per plan. This study design
can establish the relationship, suitably.

Methods ofmatching: (a) matching bypairs; and (b) Matching by mean and
standard deviation.
(a) Matching by pairs: On the basis of scores obtained, the subjects are
matched and paired. A subject who has a score of 80 is paired with another ofa
similar score on a test ofmemory. In this way, pairs are formed. Eachpair can be
seen to be a block. Another block would have pairs with scores of 90 and so on.
Here subjects with deviant scores on the matching variable would get eliminated.
(b) Matching by mean and standard deviation: Here the measures of
central tendency and variabilities in the distribution of scores on the matching
variable are used.
Three rnatching methods are employed:
(D Randomblocks methods
(D Method ofcounter balancing order and
(t'D Block-repetition method.
0 Random blocl<s method: In this method, the blocks are first created. Then
with the number ofsubjects in eachblock being kept the same, as deJermined
by the values ofthe independent variable, the subjects from each block are
randonrly assigned to the different groups as per conditions orvalues ofthe
independent variable.
l30 Seflnstuctionol Material
 ($ Method of counter balancing order: cotxfierbalancing method is used to Research Design
avoid confounding among variables. consider an experiment where the
subjects are tested on both an auditory and visual reaction time task (here
the subjects respond to an auditory stimulus) and a visual reaction time task
(in which subjects respond to a visual stimulus). If each subject is tested NOTES
first on the auditory reaction time task and second on the visual reaction
time task, then the type of task and the order of presentation would be
confounded. Ifvisual reaction time were lower, there would be dfficulty to
know ifthe reaction time to a visual stimulus is 'really' faster than to an
auditory stimuhrs or ifthe subjects learned something while performing the
auditory task that inproved their performance on the visual task. This isthe
compounding effect on the reaction time of the subject. The experiment
could be designed better using the counter-balancing procedure. Here half
the subjects would be given the auditory task, first. In this manner, the order
effects would be neutralized. So the design counter-balances the sequence
oftaskpresentation. This design enables one to conclude that the effects of
the reaction time obtained are pure and not influenced by the order of
presentation. Halfof the subjects should have been given the visual task
first and the other halfofthe subjects should have been given the auditory
task first. That way, there would have been no confounding of order of
presentation and task, as the order of presentation and task would be
'counterbalanced.'
(ll Block-repetition method: In this method, the block is first created and
then successively repeated in which the natural sequence and the same
order ofthe natural sequence ofconditions is reported for each block. For
example, ABC is repeated for each block. This means the first subjects
would be to groupA, the second to groups B and the third to group c.
The larger mean differences occur in the random blocks method, while the least
occur inthe randomblocks method.
Factorial design: when the researcher is needed to manipulate two or
more independent variables simuhaneously; the most suitable design is the fictorial
design. The benefit from this is that their independent as well as their interactive
effects onthe dependent variable can be studied. Afactorial designhas tlree main
features.
l. Two or more independent variables manipulated in all possible
cukninations.
2. Different sub-groups or subjects can serve everypossible combination
ofthe independent variables-equal number of subjects in all groups is
preferred, though this is not necessary.
3. Independent as well as interactive effects can be studied, e.g., noise and
illumination on the rate of learning. Noise and illumination are two
independent variables. Further, noise has two levers-highand low and

S elf- I n s truc ti o na I Mal e r ia I l3l

Research Design so also ilumination, high and low. The resultant fictorialdesignisa2x2.
There will be four possible combinations ofthe experimental editions.
Al B 1 for high noise high illumination
M Bl for low noise high illumination
NOTES
Al B2 for high noise low illumination
M 82 for low noise, low illumination
Then the subjects are randomly assigned to the different treatment
combinations. Inbehavioural research ergonomic frctors like age, sex,
anxiety, etc. are attributes that have to be studied. Here randomization
is to be used keeping the situational frctors in mind (women, sonBtimes
are ill-advisedto be exposedto highnoise levels).
Factorial designs with two variables 2x2 design is analysed using the
ANOVA developed by R.A. Fisher (as shown in the table below).
Table 4.5 Fictitious Data Obtqined in an Experiment Bqsed upon a 2x2 Factorial
Design

Group I (nl = 10) Group II (n2 = 10) Group III (n3 = f0) Group IV (n4 = l0)
High noise and Low noise and High noise and low Low noise and
high illumination high illumination illumination illumination low

l5 l0 l6 10
t4 t2 18 9
20 l0 22 8
22 9 25 7
l6 8 26 6
l8 t2 20 10
20 ll 20 ll
2t 10 l8 t2
l8 10 t7 l0
t7 10 t6 l0
.. X'= 181 102 198 93
..X2 = 3339 1054 4034 895
Mean= 18.1 10.2 19.8 9.3

Source: Singh, A.K., Tbsts, Measurement and Research Methods in Behavioural Sciences,
2m8.

Table 4.6: Meansfor the Four Experimental Conditions

.rl{i.$},:..i;tr:, .,
iUdr{,A) ' '.{tr4.1
,vleilE
(48,) {,+rql
E
!
.!i f.X = lEl !I - I02
I irleer.lS.l tr,L.rr- 10.2 l{.t5
O-tO nr-10
l48rl l,tQ)
a f,,t = lgE g1 - e3
I
,dcan = 19.8
J
lvkao * 9.3 Ir.ts
* l0 dr * l0
18.95 9.75
E.{ e Totil scorr on t}tc depcndmt varirble
132 Self-Instuctionol Material
 Illustration of the details of the Calculation of ANOVA in a 2 x 2 Factorial Research Design
Design (datafrom Table on the previous page)

Step h Corrcct,* * Ex' + rx' lPXt Llx.t (I8t + 102 +

!99 + 93)t
N 40

.(571f *tz:ag NOTES

{0

JtepZi Totil SS =tElrr + EXrr +IXf + IX.2)-C =63i9+ t054 + 4034 +895)-s236.9
*9322 -82369 = l0E5l

$lqp.t; Amonss$ *lILl * E{,f * QfrP * Ex,f -c

ft n2 ,] n.

_ nglr * rtgl), + rrepf * 03f _ alEes

l0 r0 I0 l0
* 32761+ l0{tt 39?0-4 + 849
- 82369 *9101! - 823L9 -8649
IO

Stqp,fr Wifi in $5 . Iual S$ - Among 5S - I 08i.l * 864.9 = 220.?

,S 5; SS between lwo lewls of A (or the first independent variable):

trxr+rxrf*E{r+u.f-c
4+nr n,+r.
*(t8t+resf
10+ lO
-OW-8236e
l( -ry-ryf -8236s

= 90833 - 82t69 = 8a&4

gp o: 55 bawecn thc two lerrelr of I (or the second tndependent variable)

G,y, + EXr)r * EX, + EX.)r _ c !l ttl o2)' (1 ?g + 9l)'

I0+lO * l0+l0
_ 82359
t\+nz nr+q =

-#. # -6236e -E236'5 -8236e=,.

Sqp 7.' lnbraction 55:

AnongSS -bttwenSSforA-betraeen SS for F= 864.9 -g/t6.4 - 1.6
*86{9 -E46 = 169

In the above table the number oftrials taken by each subject in leaming a
list of 1 5 consonant syllables with the criterion ofone perfect recitation is shown.
This data is rearranged in the next table so that the means ofthe four groups are
placed in their appropriate cell.
Then the first step is to compute the total sum of squares (ss) and then
divide it among (ss). (Step 3, next table) andwithin (SS) (step 4 ofthe table).
Before this, it is irnportant to compute the correction value (step I ofthe table).
The among (ss) indicates whether the groups differ or not. The purpose of
.
this'studywas to find out whether or not variation in each independent variable
affects the dependent variable score and whether there was any significant
interaction

Self-Instructionol Material 133

Research Design
So, the among (SS) is dividedinto three components:
(a) Between SS forA
(b) Between SS forB
NOTES
(c) InteractionAxB
In order to know whether there is a significant difference or not between
high noise conditionA, and low noise conditionA.2, the SS between these two
values is computed. Again to know the existence ofdifferenceg between high level
of illumination B 1, and low level of illumination 82, the SS between these two
values is computed. The AxB interaction is easily calculated by subtracting the
valuesofbetweenSS forAandthebetween SS forB fromamong SS (step 7).All
the numerical values from step I to step 7 must be positive. Any negative value
must be correctedbefore proceeding flirther.

Within groups design

Within groups designs are of two types 'complete within groups design'and
'inconplete withingroups design'. Acompletewithin groups design isone inwhich
practice effects are balanced by exposing the subject to the conditions repeatedly
in a different order each time. This enables the researcher to interpret the results
for each zubject. An incomplete 'within groups design' is one in which each condition
is administered to each subject only once while varying the order of administering
the conditions across the subjects in such a way that the effects ofpractice are
neutralized when the results of all the subjects are combined. The balancing
procedues for'completewithin groups design'areblock-randomizationandABBA
counter balancing while in the 'incomplete within group designs' one common
procedure isto use allpossible orders ofthe treatment and subsequent assignment
ofeach subject to one ofthe orders.
In within groups design or within subjects design, the same group ofzubjects
is treated differently in different experimental conditions and their influence on the
dependent variable is studied. This is also known as repeated-treatments design.
For example, a subject tested for non-drug and later for a drug condition. We can
then note the change in the dependent variable due to the drug intake. This type of
research is common in the fields of learning. memory reaction time and other
psychophysical studies.
The within groups design has two tlpes:
(a) Two conditions and many subjects
(b) More than two conditions and many subjects
Inthe first case, the same group is tested undertwo conditions only. The
mean difference between the two conditions is determined through a matched I
test or Sandler'sl test.
Comparison ofbetween groups design with within groups design:

134 Self-lnstructional Material

 Research Design

l. Separate groups used in experimental Same subjects used in

conditions experimanental conditions
2. No economy of subjects Economy of subjects NOTES
3. Individual differences not controlled Individual mill controlled
4. High error variance Lower error variance
5. Preparation time and effort of subjects Preparation time and effort
to be tested, high are reduced considerably
6, Can be used in a wide range of experiments Cannot be used in all types
of experiments (knowledge
of results, fatigues, drug
eff'ects, etc.)
Generallythe between groups design is considered, but more powerful within
groups desigU is preferred:
(a) when only a small number of subjects are available;
(b) when limited matching tasks exist for equating the independent variable;
(c) when convenience is a consideration; and
(d) longitudinal studies and other psychological phenomena can be studied only
by this method ofdesign.

Mixed design model

Here, one variable is fixed; the other is random
Example 4.1:
Mixed model (l fixed,8 random)
- MSfotA
Ffor-4=
MS for interaction
MSfotB
FrnrP
Lvt u -
' MS for within-groups
MS for interaction
Fiora x B = MS for within-groups
Mixed model (B fixed,l random)

FforA:ffi
MSfor B
F for B :
Ms fb, irtr-u.tio,

FforAxB:ffi
Soarce: Singh, A.K., ksts, Meastrement and Research Methods in Behavioural Sciences,
2m8.

Self-lnstructional Material l3S

Research Design Important symbols: Campbell and Stanley ( 1963) have described and discussed
16 designs ranging from the poorest to the most powerful that are useful for
psychological and educational researcher. They have also identified certain symbols
that one must be familiar with. This is a useful key for researchers.
NOTES
Fi: Random selection of subjects or random assignment of treatment to
experimentalgroups.
X Treatment or experimentalvariably, that manipulated. Whentreatments
is
are compared, they are labeled as Xl, X2 andX3 and so on.
O: Observation or measurement or test. Where there is more than one
measurement, the subscript is used O1.,OZ and 03 and on.
When X and O occur in the same row, it indicates that they are being
applied to the same persons.
When X and O occur from left to right, it indicates the temporal order.
When X and O occur in vertical order X, it indicates that the two are simultaneous.
When parallel rows of symbols are separated by dashed lines (-----) it
shows that the groups have beenunequated byrandomization.
Whenparallel rows of symbols are separated by dashed lines, it indicates
that groups have not been equated byrandomization.
Three samples of non-designs: Here, there is no control group or on
equivalentgoup.
1. One shot case study: XO is the symbolic presentation. Atreatment is
given to one group and then the observation O is made to assess the
treatment. This is a way ofdetermining the results in relation to earlier
behaviours / activities.
2. One group Pre-test-Po st-test design: Here X are judged by comparing
the pre- andpost-test scores. There is no control group present.
01 X 02 is the representation ofthis design. Attitudinal changes can be
studies bythis method.
3. Static-group or intact-group comparisons. Here, two groups are used.
One group (Ol) receives the experimentaltreatment (X) and the other
group does not (O2). Subsequently the two groups are compared.
xo1
Design ir
,OZ
The control group is used for comparing the experimental group. For groups
that not equivalent, no random assignment of subjects to the groups is made.
True experimental designs: The three designs mentioned below quahfy
as true experimental designs, because the control and experimental groups are
formed, their equivalence is established byrandomization. Relevant factors are
conholled.

136 Self-lnstructional Material

 These are the strongest types ofdesign. But under certain situations,
they Research Design
cannot be conducted.
I . Post-test only equivalent group design: This design has all the conditions
of a good erperimental design, yet it reduces the threat to the
experimental validity. NOTES
Design RX ot
RCJ2
All the other conditions ofan experiment are followed. The most suitable
statistical treatment methods are a I test orANO\A.
2. Pre-test Post-test control group design: This design is similar to the
earlier one except that a pre-test for both the groups before experimental
and control treatments are administered.
Design ROl X 02
RO3 o4
one group receives the treatment X, and the other does not. The mean
gain score ofo2 - ol, is compared with 04
-o3. The post-test means
ofo2 and 04 canalso be compared fbr ascertaining the impact ofthe
treatment on the groups.
3. Solomon four-group design: This is a combination ofthe above two
designs, i.e., the post-test only and the pre-test post-test equivalent
groups design. This controls the threats of external vatidity.
ROI x 02
RO3 04
RX05
RC)6
Here the fbur groups are randomly set by the experimenter. There is
also the advantage of reprication here, because two simultaneous
experiments are conducted. The effect oftreatment X is replicated in
fourways.
02 > 01
02>04
05>06
04>ol
This design makes it possible to ascertain the testing effects as well as
the
interaction oftcsting. with other factors. The factorial analysis ofvariance
is a
suitable statisti*r i method.

4.3.2 Quasi-Experimental Designs

This is a situation where an experimental interpretation is applied to results

that do
not meet the criteria of a true experiment. There is some control over the
manipulation ofthe independent variable but not in creating the basic equivalent

Self-lnstructional Mqterial 137

Research Desigtt groups. This is a means to simulate a true experiment. So it is called a compromise-
design. These are like experirxents bccause the control group is used. But the
subjects are not assigned randomly, so the validity ofthe experinrent is questionable.
The in{ependent variable is not trulymanipulated. So, the studyis called a quasi-
NOTES experiment. Subjects are assigned to a particular condition that already exists like
age, sex, personalitytraits IQ, etc.

These designs lie between pre-experimental and true experimental design.

Time-series design: The design is
o1o2.o3C,4 x 05060708
Here a series ofpre-test are administere<lto the group. Later the treatment
X is given followed by a series ofpost-tests ali for the satne grotlp. Extraneous
variablcs are controlled here. The background variable is not controlled as the
subjects are exposed to different stimulations daily. The results are analyzed by
simple statistical comparisons. O4 and 05 comparisons are avoided. All other
pairs are compared.
Equivalent time-samples design: It is like the time-series design. Asingle
group is used and the group is adrninistered repeated treatments in some systematic
way.
Designs is
x. 01, xo 02 x103 xo 04
The treatment X I is irrtroduced and re-introduced rvith sorne other experience
XO which is presented in the absence of the treatment. The background history is
well controlied in this experiment bypresenting X several times. so the external
rariables get well-controlled. This design is an improvelxent ol'er the time series
desrgn.

Non-equivalent control-groups design:'fhis is useful rvhen the researcher

has to work w.ith intact groups rvirose mernbership is alreadyexisting (e.g., students
offirst year ofcollege) the subjects cannot bc reconstituted for some administrative
reasons. In dealing rvith intact groups the design is

01 x02
o3 o4
This shows that the random assignments of subjects has not been possible
to the experimental and control conditions. So the equivalency is absent' The
intact groups are compared on the pre-tests O I and O3. The statistical analysis
involves comparing the mean gain score ofthe treatment group (O2-O I ) to that
ofthe non-treatment group (O4-O3). This non-equivalent group design cannot
be used rvhen tlre intact groups are dissimilar.

Counterbalanced design: Here the experimental control is achieved by

randomly applying the experimental treatments. This design is also known as cross-
over design or switch-over design and rotation experiments. For counter balancing
138 Self-lnsnactional Material
 each treatment appears only once in each column and in each row. A counter Research Design
balancing design involving lour treatments and fbur groups on four different
occasions is given belorv:
GroupsA XIO X2O X3O X4O
NOTES
Groups B X2O X4O XlO X3O
GroupsC X3O XlO X4O X2O
Group, D X4O X3O X3O Xl O
Almost all the variables threratening the intemal validity are well controlled.
The only limitation ofthis design is that of multiple X interforence. This reduces
generalizability.

Separate sample pre-test-post-test design: This design is used when

the experimenter cannot assign treatments to all subjects at the same time. So, a
sample is choose'r to receive the treatment. Then another sample is chosen and
given the same treatment. For example, 5000 newly recruited persons are to be
trained by the police. only 200 persons can be tarried at a time and all people
have to be trained. S0 the division ofpeople into training conditions and non-
training conditions does not occur. So, one group is chosen as the pre-test post-
test one, and another group for repetition to enable comparison

- post-test
]

Design separate sanrple Pre-test i

olx02
03x04
If 02 exceeds ol and 04 exceeds o3, we can conclude the effect ofthe
treatment X. Since the two groups are separate and the independent variable is
administered, some amount ofpower to the conclusion is enabled.
Patched-up design: Here the experin-renter starts with an inadequate design
and adds more features, as one proceeds.
GroupsA X Ol
Groups B 02 xo3
This isused to overcome the weakness ofa given design. The comparisons
made are between 03 and 02 and o2 and o I . The groups are tested in slquence
and then compared.

Quasi-independent variable oftime is a r,r,ellstudied variable. Develgpmental

psychology uses this externively.
For the study o ftirne. three designs arc used:
l. Longituilinaldesign
2. Cross-sectional design
3. Cohort design

Self-lnstruclional Material 139

Research Design Longitudinal design: Here a group of subjects are studied to see the
effect of time. There are a lot of confounding factors present here. Motor
development, verbal developments etc., are all examples of this. Repeated
observations ofeach subject at different ages are carried out.
NOTES
Subjectsl 5 6 7 8 9 10
Subjects2 X X XXXX
Subjects3 X X XXXX
Subjects4 X X X XXX
x x x xxx
X and X show the scores and means. Adding vertically the mean scores for
the conditions reflects the changes occurring as a function ofage.
Advantages of longitudinal study:
1. Subject variables are held constant between the conditions, e.g., genetic
factors
2. Sample equivalence problems do not arise
3. Cause--effect connection established, well
4. lt cantrace growth increments and patterns
Disadvantages:
1. Extraneous variables come into play
2. Societal and cultural factors do influence
3. Due to repeated measures, carry over effects may occur.
4. Environment ofthe subjects cannot be controlledbetween testing periods

5. The researchers get stuck into an early determined design.

6. Expensive in terms of time and cost

Cross-sectional study: This is a form of between subjects quasi-
experiment. The researcher observes the subjects at different ages or at different
points in temporal sequence. A cross-section of ages is selected. For example,
studying vocabulary a group ofsubjects of5 year olds, another of6 year olds and
so on are observed at the same time. Then conrparisons are made:

5678910years
x xx xx x
x xx xx x
x xx xx x
x xx xx x
xx xxxx
X represents the scores and X represents the mean ofdifferent age groups.

140 Self-lnstructional Malerial

 Advantages: Research Design

o Cross-sectional study can be conducted quickly and easily

o It iscost effective
o Data are collected at one go.
NOTES
Disadvantages:
o Directional change in not indicated bythe study
o Ckonologrcal age is the same, but maturitylevel, not controlled
o Developmental milestones ignored
o Comparability of groups, unsure
It is a form of snap-shot study
cohort design: Here, longitudinal studies ofseveral groups are done and
this is usually a form of gerentological study or developmental study.
See figure 4.6

Diagram ofa cohort study

Repeated measures over ages
t Diagram of a cohort stu{
f, epeattd mea$ulcs ov*r ate$

5 a
D B 9 IO
1990 r x x x x x x x
Subiects z x x
3 i i i x

I 995 I x x x x x x ;erreralion
Subjectr x
2 X x x x lmain effect
: :
20CI0 'r
x X x X A x x
Subjecu ! x x x x x X
3 :

matn
Age ma X xxxx
effect

Fig.4.6 Cohort Design

Source: Singh, A.K., Tbsts, Measurement and Research Methods in Behavioural Sciences,
2008.

The columns represent the repeated measures over the age. The rows denote
three respective generations. Age effects and generation effects, can both be
ascertained
Ex post facto design: Here the experimenter does not introduce a
treatment, but evaluates a naturallypresent treatment, or one that has already
occurred. The dependent variable is related to the conditions that already exist.
Two types ofcommon expost facto designs
l. Correlationaldesign
2. Criterion-group design

Self-lnslructional Material l4l

Research Design Correlational Design: This is also called a psychometric approach. Here
two scts or more ofdata are collected from the same group of subjects so that the
relationship between the two subsequent sets of data can be determined. The
correlations form the main statistics for analyzing the data. Other variants of
NOTES corelational techniques can also be utilized.
Diagram ol'design
ol 02
Stucl v*: To examine the relationship betrn'een anxiety and performance of a
group ofchildren ofClass XII. First a measure ofarxiety is obtained and later a
test ofperforrnance. So. two sets ofdata are arailable--Ol and 02. The com ''lation
coefficient can be calculated. It can be positivc. ncgative or absent. Correlations
only signifo the nragnitude ofthe relationship betu,een the two variables. It is not a
cau se-effect relationship between tlte variables.

Criterion group design: I{ere groups with contrasting characteristics are

selected and the experimenter tries to ascertain what has caused the particular
state by examining groups which have certain criterion u,ith those which do not
posses it.
Diagrarn ofthe design
c ol 01 cC)20l
or ____
02 03 04 02
C-selected criterion: It is a kind ofstatic group conrparison. For exanrple,
in a study ofcreativity in children, 50 students are selected randomly. Then, a test
ofcreativityis administered. Then the stude:rts arc divided into two groups-high
and low on creatiyity. Later the experimeuier tnes to relate this to the child rearing
practices in the farnily. This can be done by the struoturcd inter ,*-ierv methods. The
results could point to a set of potentiai factors leading to creativity, but not a
definitive conclusion.
Multivariate analysis: ln rnodern times, rvith the availability of the
computer and variotrs types of software packages, several multi'uariate analyses
have become possible to treat data. Multivariate analysis includes all statistical
methods which can simultaneouslir analyze more than two variables on a sample
ofobservations.
Multiple regression anall'sis: This analysis is suitable when there is a
dependent variable that is influenced bytu'o or lnore independent variables. In a
simple regressionmodel, the dependent variable is a function ofone independent
variable, e.g., sales as a function of advertisernent expenditure. Suppose the
regression yields ft2 :.60 and the regression coc-llcient is significant. This means
only 60 per cent ofthe variation has been explained. Thc other.tl0 pr cent is due to
some other influence besitlcs expcnditure. So. rnore explanatoryr,ariables need to
be added. One such variable may be the per capita inconre of the trading area.
The regression anall,sis that uses two or rnore independent variables is termed
multiple regression analysis.
112 Sell-lnstruc'tionul -llatcriul
 Multiple discrirninant analysis: This method is useful when there is a single Research Design

dependent variable that cannot be measured, but can be elassified into two or
more groups on the basis ofattribute. The object ofthis analysis is to be able to
predict that a particular entity belongs to a particular group, based on several
predictor variables. NOTES
MultipleANOVA: This is an extension ofthe trvo-u,ayANOVA.
Canonical analysis: This is a useful rnethod when both the measurable
and non-measurable variables are present. This method can help in simultaneously
predicting a set of dependent variables fiomtheir joint covariance rvitha set of
independent variables.
Inferential analysis: This is concemed with various tests of significance
for hSpotlresis testing. It can also help in estimation ofpopulation values. Conclusions
are based onthis type ofanalysis.

CHscxYoun Pnocnrss
5. What is matched groups design also known as?
6. Name the different types ofmatching.
7 . S,hen is a factorial design used?
8. Definetrueexperimentaldesigns.
9 " What is a correlational design'/
10. What isthe cohorl-design?

4.4 QUALTTATIVE METHODS

Qualitatire research is gaining a lot ofimportance in the socialscienccs. There are

specific methods u'ith dillerent assumptions and objectives. These are intrinsic to
all types ofresearch. The steps ofthe research process and the proccdures for
collecting data and analyzngand interpreting it are different.
Today, inductive strategies are incrcasinglyrrsed in research instead ofstarting
frorn the basics and testing them crnpirically. Traditionally psychology and other
social sciences have used the tnodel presented by the natural sciences for developing
quantitative and standardized methods. The guiding principles hare *ken idcntifyng
the causes and effects and the measures have been quantity phenomerla. These
have pemritted generalizations and the forrnulation oflaws. Causal relations are
studied under controlled conditions and obsen ations have been classified in tems
oftheir frequency and distribution. Studies have been made as objective as possible.
For a long tirne, psychological research has depended on experimented designs
for study as the rnainstay.

S?ll-lnstructionql Material 143

 Research Design
Theresults ofsocialsciences are not applied indailylife as the methodological
precisions are far too removed from every day problems and situations. The goal
of subject and situation-related research can be achieved with qualitative research
inameaningfulway.
NOTES
It is important to recognize that not all questions ofresearch can be studied
empirically. Sometimes methods are not available for stuclypurposes. Complex
social science problems are not easy to be put into the mould of empirical
investigations. It is not easy to reduce complex situations to simple and single
variables for the sake of studying. The entire complex situations are rendered
rather simplistic and therefore, often inappropriate" The subjectivities of the
researcher and those being studied are part ofthe qualitative research process.
The qualitative resemchers are soft, rurderstanding, ope& and descriptive strategies.
This is in opposition to the hard experimental, standardungand quantifying
approaches.

4.4.1 Definitions of Qualitative Research

It is a method ofinquiry used in many different academic disciplines, traditionally
inthe social sciences. It is a set ofresearchtechniques in which data is obtamed
from a relatively small group of respondents and not analyzed with statistical
techniques.

Qualitative research follows an inductive research process and involves the

collectionandanalysis ofqualitative (i.e, non-numerical) datato searchforpatterns,
themes, and holistic features.
There is no formal quantitative framework used to generate projections in
qualitative research. It is concerned with understanding the processes, which
underlie various behavioural pattems.'Qualitative' is primarily concerned with the
'why'ofinvestigations. case study, observation and ethnography are considered
forms ofqualitative research. Results are not usually considered generalizable, but
are often transferable. A subjective form of research that relies on analysis of
controlled observations ofthe researcher. This type ofresearch yields extensive
narrative data, which include detailed descriptions ofwhat has been observed.
Here, collection ofnon-numerical data using interviews, observations and
open-ended questions, to gather meaning from non-quantified narrative inlbrmation,
are involved. It is a form ofresearch that derives data from obserr,,ations, interviews
or verbal interactions and focuses on the meanings and interpretations ofthe
participants.

This is a way to study people or systems by interacting with and observing

the subjects regularly. It is a free-form research technique that is used to gain
insight into the under$ing issues surrounding a researchproblem bygathering
non-statistical feedback and opinions rooted in people's feelings, attitudes,
motivations, values and perceptions, often from small samples, also called soft
data.

144 Sel.f-Instructional Materiul

 The gathering and analprng ofdata based on interviewees, own perceptions Research Design
or experiences in order to provide insight into their beliefr about their circumsiances
rather than measurable data. Subjective information is obtained from groups
and
in-depth inten'iews and participant observation.
Qualitative research is generally
conducted on a small group. It looks at phenomena like attitudes, perceptions, NOTES
etc.

4.4.2 Distinctions between euaritative and euantitative Research

Qualitative methods are used for exploration, especiallyfor generating

fu4pothesis.
Quantitativemethods areused forhpothesis testing. Establishingthe content validity
ofanymeasure is the strength ofthe qualitative research.
Qualitative research is
used for evaluating anypolicy, program, etc., more effectivelythan quantitative
research. The how and why ofphenomena are better understood byqualitative
research. Certain questions like relevance ofa particular topic is better studied
qualitatively. This type of research allows for more diversity in responses and
it
can adapt to new developments, as theyhappen. The count is not important
as in
quantitative research, but the material got through study is seen as valuable.
The aim of qualitative research: The primary aim ofqualitative research
is to gain insights into anyphenomena of interest to the res.urch"..
For example,
one is interested in studying mental disorders. Epidemiological studies show
the
frequency ofschizophrenia and the distribution of this disorder in a population.
The finding is it is higher in the lower socio-economic classes. These correlations
have been well-established by empirical quantitativ e data.However, it is
not clear
whether the conditions of living in the lower economic conditions foster the
development of schizoplu'enia or people with schizopluenia get marginal ized
and,
drop into the lower economic strata ofsociety. Often studies failto capture the
contextualperspective ofthe illness. What is the subjective experience ofbeing
schizopltenic? How do the family members deal with this illness? what are the
benefits ofinstitutionaltzngsuch a patient? How do professionals view women
with schizophrenia? These are some ofthe relevant questions that qualitative
research seeks to find answers to.
For qualitative research. the mental illness starts with the patient, his / her
relatives and the professionals involved and their perspectives. tt also analyses
the
interactions in dealing with the illness for all concemed. The different viewpoints
are significant in this type ofresearch. Communications between, the investigator,
patients, psychiatrists and others in the field, are part ofthe knowledge
about the
disease. The subjectivities ofthe investigator a well as that ofthe patient,
constitute
part ofthe research process. The researcher's reflections frusirations,
feelings,
impressions and so on all become part of legitimate data in their own right
in ihe
context ofdealing with patients involving the illness.

Qualitative research is :

o Areturn to narratives, language andcommunication

o Questions about the specific, particular and concrete problems
Self-Instructional Moterial 145
Research Design . A study ofthe local contexts, instead ofuniversals
o Putting problems to be studied in their temporal, historical context and
descnbe and explain them are different approaches to qualitative rcsearch.

The approaches have different theoretical assumptions:

NOTES
L Subjective meaning: This is the meaning individuals attribute to their
activities and environments.
2. This meaning is derived from and alises out ofsocial interaction that one
has withones fellows.
3. These meanings are handled and moditied through interpretations by
people while dealing with r,'arious interactions.
'The researcher has to see the world fromthe angle ofthe subject, he/she
studies'(1oulmin, 1990). This means that methodologically the researcher has to
reconstruct the subject's view point in rarious ways. This is fur the form of subjective
theories people have in explaining different aspects ofthe world. For example,
The role ofpunishnrent in disciplining children; How the unir,erse was for med?,
etc. Herc autobiographicalnarratives, biographical reports. etc., are all significant,
as methods.
IVaklng ofsocial reality: Ethno methodology is one ofthe techniques used
for studying how people construct their own social reality. Garfunkel (1987)
proposed tl'ris rnethod. Here, every day activities are analyzed in terms oftheir
social context as to why some activities occur everyday. All these are viewed as
visibly rational and reportabie for all practical purposes. This research is based on
conversation analysis.
Conversation analysis is based on three basic assumptions:
1. Interaction is structurally organized
2. Theyare context shaped and context renewed
3. Conversational details cannot be dismissed as disorderly or irrelevant.
Cuhtral framing ofsocial and subjective reality and cultural systems determine
the perception and making of social reality.

4.4.3 Research Perspective

Study oi subjcct A and

Anall'sis of intclaction Stud1,'olsubjcst B and
or her r ieupoint(s)
hi-s

t-z
his or her vier.r,point(s) and discourses

Stud-ving cultural lianring of practiccs

Fig 4.7 Researc'h Perspective,s in Qualitalive Research

Source: Flick, An Inlroduction to Quantitatit'e Research,1998.

146 Se If- I nslruc lion a I L{a terio I

 A social representation is understood as a system of values, ideas and Reseorch Design

practices with a twofold function: First to establish an order which will enable
individuals to orient themselves properly in their material and social world and to
master it; and secondly to enable communication to take place among the members
of a community by providing them with a code for social exchange and a code for NOTES
nanring and classifying unarnbiguously the various aspects of their world and their
individual and group hlstory"

Qualitative research aims at understanding an event from the inside. It is the

view of one subject or subject(s), the social situation or the cultural or social rules
relevant for a situation that have to be understood. Asingle case is analyzed before
a general statement is made.
L First a single subject theory
2. Then a single conversation
3. And then a single cases is reconstructed
4. Later other case studies are used as a comparison.
5. Then a typology or different subjective theories are developed
The reality studied by qualitative research is not a given reality, but as
constructed by different 'actors'. How crucial an actor is, is determined by the
theoretical position adopted to study the process ofconstruction.
In the process ofreconstruction, texts are used for empirical analysis. The
view ofthe subiect is presented as his/her subjective theory. For example, beliefin
astrology where the course of interaction is recorded and transcribed and then the
reconstntctions oflatent structures ofnranning are formulated fromthe texts (i.e.,
what is toldandthe actions that follow, e.g.. matching ofhoroscopes). The status
of the texts will be detennined by the theoretical position held.

Table 4.7 Theoretical Position in Qualitative Resear.ch

Subjects' points of Nlaking ofsocial Cultural framing of

vierv realities social realities
Traditional theoretical Symbolic Ethnomethodology Structuralism,
background interactionism psycholanalysis
Recent development in Interpretive Studies of work Poststructuralism
social sciences interactionism
Recent developments Rcsearch Discursivc- Social represcntations
in psycholog.v programme psychology
'subjective theories'
Common features o Verstehen as epistemological principle
. Reconslmcting cases as starting point
r Construction ofreality as basis
o Text as enpirical material

Source: Flick, An Introduction to Quontitative Research,1998.

Self-lnsrntt tional Materiui 147

Research Design
4.4.4 Texts and Realities in Qualitative Research
Texts serve three purposes in qualitative research.
l. They are not only the essential data, but the basis for interpreting and
NOTES cornrnunicating the findings
2. Texts are seen as an instrument for interpretation ofsocial reality
3. Interviews comprise the data that are transformed into texts.
Hereafter the text is used as a substitute for the reality
Text making ofthe world involves first degree and second degree constructions.
Realities are activelyproduced by the participants through the meanings ascribed
to certain events and objects. This is what qualitative researchers seek to study.
These ideas ofsocial events in a social field may conpete, conflict or succeed and
are shared and taken for real. This is how reality is constructed.

Construction Texts as versions of

the world
/\,/ \
/\,/\
r---\ \\
Experience
\
lnterpretation

Natural and social Understanding,

environment ascrlption of
events, activiiies me'aning

Fig.4.8 Realities

Source: Flick, An Introduction to Quantitative Research, 1998.

Strictly speaking there are no such things as facts, pure and simple. All facts
are from the outset selected from a universal context by the activities ofour mind.
Theyare, therefore, always interpreted facts, either facts looked at as detached
fromtheircontext byan artificialabstractionor facts considered intheirparticular
setting. ln either case, they carrytheir interpretational inner and outer horizons.
For Goodman (1978), the world is sociallyconstructed through different
forms ofknowledge-from everydayknowledge to science and art as ways of
making the world. Social research is an analysis of such ways ofworld making
and the constructive efforts ofthe participants in everyday life. Acentral idea in
this context is the distinction between first degree and second degree
constructions-first degree isthe construct madebyan actor. The constructs of
the constructs made by the actor in the social scene is the second degree. The
exploration of the general principles, according to which man organizes his
experiences in daily life and those of the social world, is the first task of the
methodology of social sciences.
So, there couldbe muhiple social realities. Social science research encounters
f48 Self-lnstructional Material the world it wishes to study only in those versions as constructions by the subjects.
So, there are subjective constructions by the participants and subjective Research Design

constructions bythe researchers. Knowledge ofthe world is not just found in the
world, rather it is built into it. worlds are made from other worlds. Abig part of
research involves reconstructing life stories or biographies in interviews
NOTES
4.4.5 Theories in Qualitative Research
To beginwith, qualitative research is circular, not linear like in quantitative research.
In this tlpe ofresearch, theories undergo revisions, evaluation, construction and
reconstruction. They are versions ofperspectives through which the world is seen.

T-_
tpI
Idt
-1
[t
loi
I:I
t>1
II i*:_l
t
F---e*'r co
Int F
lal 6
l9l E
iEl
ict 6
c

t
cl
o,
ut
et
al
U
()I
I AJ
!?

{
\€od
,lo\
'o\
O\*
\S
.U
*{
..1o
\s \
qa!)

q)

o)

(-)

\
r.

(.)

lrt
p

Self-Instructional Material 149

Researclt Design
Each formulation has its own perspective hidden within it. The perception in tum
feeds back into the social construction ofthis perspective and thereafter the world
aroturd us. So, theories are mere prelirninaries and relative. When the version is
developed further additional interpretations become possible ofthe new material
NOTES and so on. So, there is a pre-understanding from which, all study starts. Glasser
and Strauss suggest a circular model for qualitative research.

Formulating research question in qualitative research

R.esearch questions exist in the investigators personal biography and social context.
Certain issues are brought to the fore and others igpored.

Fcrr:,;i.!lr*fl {f th* r}\ttrell

rl iF!l:17

*
'rrlljitirilli ti :il,.licti,r: resga
ilLreglii:l!

t
;ft$i jliii.or di lor*,?irrr*
i:,)11i-:iriliS

t I

S*i*cii** *f r*selri:l gl*,.;ns'.n":h

*brir, 11 .',:!r'J! 1l,l al'r.-]$llrla
i

lrai*eii;:t *r*: lgi,lrmu:iitr{,il qi

tle spec,ii+ iesFirrai qre!'rirli:
I
Cclie{:i*n
^t a-r^
i
t
Lrirlualiiri <1i!j .*f+r,alirl*iia,n ri
1te lps*ilit resesfcn or.;*si:ilns
I

f
AraiY*irg
?he ilsis
I
t
6et-re:'alilalel and glrii;il!-l ci
tFr$ anBlYSEr
I

i
r,;rIogi111gr. g;

tre i;lrd:r"es

Fig. 4.10 Reseurch Que.stions in the Researc'h Process

150 SeUtlnstrttLtional Material

 The research questions can be of two types: Those oriented towards Research Design

descnbing states (whichkind, howoftq etc) and those orientedtowards describing

process (causes, strategies, etc)

NOTES

.iEB

€E
el
Eg

l! 3
a,
c,
o

it
E
-
6

=*
c
BS E
e
E $- E
a
6 tE 3
t)
B6 C
* e
-8
E E
3 .,
It cod
G
0
I
?e
-1
E.S
EE ES
tr-
lE^:

Pg E:
C=
5
E
o
'L

gH {E
s3
4S
3
r
tt{l
E
t o.s
FS

E:
F
I Ir
a!
EA
3S
o
I
E
E:
a\
!
F C
aCJ -sz

oq t-
Ei
st J
l{
a
F f!$Hn$HEE E$ f;s
Self-lntn ,"rrorr, Material l5l
Research Design
Aresearcher can enter the field ofstudy asr a stranger, visitor or as an insider.
The best role to adopt is that ofan insider. The set ofrealities presented
would be most similar in the role ofan insider and ideal for qualitative research.

NOTES 4.4.6 Sampling Strategies in Qualitative Research

Instead ofselecting a sample, the complete collection method is used in qualities
research. Thesample ispre-determinedbycertaincriteria, e.g., acertaindisease,
age, gender, region, marital stafus, etc.

Sample is also defined gradually. They are made on the basis ofthe groups
to be compared or may focus on specific persons. The sample is chosen on the
basis of new insights for developing a theory. Groups or persons chosen for the
sample are stopped when saturation is reached, i.e., nothing new could emerge
hereafter.

Samples of gradual selection

Table 4.9 Theoretical Versus Statstical Sampling

fh.adG.l r.nffng ttrdrdcd sunplng

E$rrim of fhe baoc poguhtim b rm E (Errlon ol trG he.r popuhtbn B known
lsuat h adrrrn in advancr
Fcatrrrs d the bark populsdon arG not Dist'lb.nbfi sf feerEs in fia balic
krEwn ln adEncc populat'ror cll be €ttimrtad
neprted drawing of Impling iderrlants On+*ot tlrarv{ng of a ramflc fdlowlng a
with critd! to be dellncd again ln cedr plea de{lned In aclvance
itGD
Samplc ilch not defued ln .drance Sempk da b dellned in adnance
Sampntlg h ltnbtled uttGn treoGtk l Sampllrg h fini*rd tytren the r,*rola
rdrrtio.r has brcn rudrcd sarpb ho bcen studied
Sourc Sriddnsnn t9,91, p.llt
Source: Flick, An Introdtrctiort to Quantitative Research,1998.

Sampling proceeds according to relevance, not representations; and witlth

and depth as the aim of a sample, not representativeness.

Table 4.10 Sampling Strategies

Sampling strategies in qualitative research

. A priori determination
o Complete collection
o Theoretical sampling
o Extreme case sarnpling
r Typical case sampling
o Maximal variation sampling
o Intensity sampling
o Critical case sampling
o Sensitive case sampling
o Convenience sampling
o Primary selection
o Secondary selection
S o urc e : Flick, An Int roduc ti on to Qu ant i ta t it e R e s e arc h, 1998.
152 Self-lnstructionalMaterial
 The object ofthese sampling strategies is to provide richness ofinforrnation. Research Design
It also helps in obtaining relevant information. It is an intense attempt at data
sanpling. Interviews shouldbenon-directive, qpecific, average, in-depthandshould
have personalcontext and content.

Sample selection is based on the principle ofgradual selection. It is alway,s NOTES

purposive sampling, not random sampling. Palton gives the following concrete
suggestions:
o Integrate purposivelyextreme or deviant cases. Cases ofsuccesses or
faihres are chosen and anallzed. This is thought to help inundostanding
the whole field.
o Another feature is to select typical cases. These are the average for the
group. This means looking at the field from the centre and from the
inside.
r To choose a maximal variation sample-here, the idea is to integrate
only a few cases, but which are as different as possible. This is aimed at
obtaining the range ofvariation and differentiationina field.
o choosing cases onthe basis ofintensity-here, the expectation is that
the interesting features, processes and experiences that go with this would
be integrated inthe studyand compared.
o Selection ofcritical cases is done to enhance the firnctioning ofthe study,
by virtue ofits richness.
o Choice ofsensitive cases-these are to be inchrded due to their explosive
force and, therefore, the unique prospect of studyingthemmaking it
special.
o Finally, the concept ofconvenience in selection-these are included
because ofeasyaccess, reduced effort and decreasedtime whenpeople,
money, etc., are limited.

Cases are considered as samples

Each case is thought to have five representative aspects:
(i) A single case is understood as an 'individualized universal'
It is viewed as a
result ofspecific individual socialization against the general background,
e.g., a physician or psychologist trying to understand a disorder. Such a
socialization can be understood in the social context, leading to different
subjective opinions, attitudes, etc. This gets to be displayed inthe actual
interview situation
(D To understand the 'individualized universal', the case is seen in a specific
context in whichthe individual acts and which he/she represents to others.
So, the doctor or social worker orients his or her practices to the institutions
under which these dfficulties arise, E. 8., ?dysfunctional family, night shift
work, etc.

Self- In*ruaiona I Material 153

Research Design (O An individual case is also seen as a specific form ofprofessionalisation,
e.g., information engineer, social worker, etc. Therefore, the training ofthe
professional individual and the actions that arise from that context got to be
studied and analyzed. Therefore, different professionals approach the same
NOTES problem differently. This urakes a given case a unique sample.
(iv) The case also develops its subjectivity because of the specific stock of
knowledge and the corresponding ways ofacting and perceiving come into
play.
(v) The case also represents an interactive mode ofcontext and activity.
Sarnpling decisions cannot be rriade in isolation. The appropriateness ofthe
content and the strategy together determine the selection. It rnaybe difficult to
make generallyvalid statements on the basis of a single study, but it is also equally
diflicult to give deep descriptions and explanations ofa case based on the principles
ofrandorn sarnpling. This is the strongest argument ofthe qualitative researcher.
Since sampling strategies are meant to disclose a given field, it can start fromthe
extretne, tlre negative. the critical and the deviant cases and thereafter form the
extremities ofthe field. It can also starl ltom the inside, which is the case ofthe
typical orthe developed cases. There, sampling can start fromthose cases which
are as far different as possible in their variation. This precludes the need for
homogeneity in the sampling. Sampling decisions determine what becomes the
enrpirical material in the form oftexts and what is taken from the texts and how it
is used.

4.4.7 Different Types of Interviews for Qualitative Research

4.4.7.1 The lbcused interview
First part: It has turstructured questions, like what did you like most about?
Seconcl pan.' Serni-structured questions-they are: left open-ended for being
able to get a feelabout a given rule.

Thirul part: Structured questions like was the lecture intbrmative?

4.4.7.2 Problem-centred interview
It involves asking questions like 'How open do you show the allergy?'
Object-orientation to detennine cr problent: Do you see yourselfas healthy
orunhealthy?
Process orientation: Do you search the Net to know more about your
allergy'?

4.4.7.3 Expert interview

This type ofinterview is directive, stronger and guided.

4.4.7.4 Ethnographic intervierv

Here the project is explained in Cescriptive and structural questions.
l5{ S.'ll-lnstil(itndt l,l.tteriul
4.4.7.5 Narrative interview Research Design

Here knowledge and experiences are presented. This is called a generative nanztive.
This isto obtain or elicit answers on a theme ofstudy.

4.4.7.6 Episodic interview NOTES

Here episodes are presented as a source ofknowledge interms ofcircumstances
or events. Then the semantic knowledge (the meaning) attached to the events is
studied through the narrat ion.

4.4.7.7 Group discussions

Observing the responses ofa small group at one time is called a group discussion.
An explanation is given about a topic, procedure, etc. Members are introduced as
a warm up act and then asked to discussion the topic. The opinions givenbythe
mcm.,.,hrs are got from verbal data. Statements and thoughts are gathered. This is
a dynamic process in operation.

4.4.8 Focus Groups

Here the focus is on the interaction among members. A focus group is a form of
qualitative research in which a group ofpeople are asked about their perceptions,
opinions, belieft and attitudes towards aproduct, serice, concept, advertisanrnt,
idea, or packaging.

All the above methods are used for collecting verbal data. The method to
be chosen is on the basis of its appropriateness.

4.4.9 Visual Data

4.4,9,1 Observation
Observations can be ofthe following types.
o Coveft vs. ovL'rt observation
o Non-paft iciparrt vs. Participant observation
o Systematic vs. unsystematic observation
o Natural vs. artificial situation observation
o Self-observation vs. obsen ing others

4.4.9.2 Phases of 0bservation

Authors likeAdler andAdler ( 1994), Denzin ( 1989b) and Spradley ( 1980) name
the phases ofan observation as:

The selection ofa setting, i.e. where and wheri the interesting processes and
persons canbe observed;
The definition ofwhat is to be documented in the observation and in every
case;

Self-lnstnr<'ti,nul Ltaarial 155

Research Design The training ofthe observers in order to standardize such focuses;
Descriptive observations which provide an initial generalpresentation of
the field;

NOTES Focused observations which concentrate more and more on aspects that
are relevant to the research question;

Selective observations which are intended to purposively grasp only central

aspects;
The end ofthe observation, when theoretical saturation has been reached
(Glaser and Strauss ,1967), i.e., further observations do not provide any fresh
knowledge.

4.4.10 Ethnography

The ethnographer participates, overtly or covertly, in people's daily lives for an

extended period oftime, watching what happens, listening to what is said, asking
questions. He collects whatever data are available to tluow light on the issues with
which he or she is concerned. He may show photos to the interviewee while
interviewing andsometirnes he mayuse flhnanalysis questions ormicro analysis of
claps done.

4.4.11 Coding and Categorizing

The interpretation ofdata lies at the core ofqualitative research. For coding and
categoizngare crucial. The texts obtained can either be augmented by alteration
or paraphrased, summarised or categorized.

The process of coding

There are three types ofcoding:
(i) Opencoding
(if Axialcodingand
(ii| Selective coding
These are ways ofhandling the textualmaterial. Coding is the process by
which data are broken down, conceptualized and put back together. The coding
is initially is done closely to the text as possible and then made more abstract.
Categoinngrefers to the summarizing ofsuch concepts. Another wayofcoding is
by themes. These are useful for comparisons.

4.4.12 Content Analysis

This is a classical procedure for analyzing textual material This includes both visual
and interview data. The categorizing done earlier a.re used and modified, as required.

First step is to select the relevant parts of a text;

Second is to analyze the data with who, what, when, whom, how etc.,
questions.
156 Self-lnstmctional Material
 The coding unit presents the smallest element. The categories
- - provide Research Design

the largerunit

The contextualunit determines the largest elements for analysis.

Techniques of content analysis NOTES

Three basic techniques are:
1. Paraphrasing the material--here the less relevant details are skipped (first
reduction)
2. Bundling and summarizing-similar paraphrases are grouped together
(second reduction)
3. Generalization-summarizing at higher level of abstraction.

4.4.12.1 Sequential analyses

In order to analyze the elements or statements, it is necessary to put things
contextually. Here sequential ana$sis is useful. Any idea is put in social order,
which also gives the understanding ofthe interactions that occur. So, the contents
are presented in a reliable way in the best possible context.

Conversations, discourses, narratives are all interpreted using sequential

analyses.

Making qualitative research reliable and valid: The legitimacyofthis

type ofresearch has always been criticized. One ofthe ways to achieve reliability
is to train the observers before they enter the field and evaluate the observations.
This is to improve comparability. Certain conventionalized procedures for field
notes are an exarple of increasing reliability.

Table 4.11 Symbolsfor Reliability

Sign Convention Use

Double quotation marks Verbatim quotes
Single quotation marks Paraphrases
() Parentheses Contextual data or fi eldworker's interpretations
Angled brackets Emic concepts (of the member)
Slash Etic concepts (ofthe researcher)
Solid line Beginning or end of a segment

Source: Adapted from KirkandMiller 1986; Silverman,1993.

Validity: This has received more attention thanreliability. Certain guidelines

have beenproposedwith this in mind.

4.4.13 Documentation of Data in Qualitative Research

For interview data-the spoken words are edited and transcribed. For
observational data-the documentation involves recording the actions and
interactions. In both these the contextual enrichment ofthe data is an important
aspect ofthe documentationprocess. The procedures involve the texts, which
becomes the basis for analyses. There are four steps in documenting data. Self-lnstructional Material 157
Rcscurch Design (i) Recording the data
(ii) Editing the data
(iii) Constructing a new reality in and from thc produced text.
NOTES (iv) This is thc way in which the construction of'rcality in qualitative research
takcs placc.
Audio-visual acoustic and visual recordings are all part ofthe new recording
possibilities ofdata collectfurn. Howcver, what is worrisome is that thc prcscncc of
thc recorder and the recording instruments could affect the recording. I'lcrc thc
least intrusivc ofequiprnent are to bc used and thc rninimal prescnce ofpcople is
suggcsted in thc observation areas.
For field notcs, the qualitativc rcsearcher is told to take notcs and record
the observations ilnn'rediatcly. Sincc the rcscarcher is often one who partakcs in
action research to a largc extent. it is also suggcsted that notes be
complctcd alier
ending the study. The rcscarclrcr should spend as rnuch timc on noting the
observations as was spent for carrying out the observation--the general nrle of
thumb for effective documentation. This is how reality is to be produced from the
field notes. Along with this the protocols ofdiaries should also be rnaintaincd. for
corroboration.

Field notes documentation

o Clondensc conversations into single words, sentcnces etc.
o Expand thc anecdotes from interviews and observations
o The field -- work journal must have an account ofexperienccs, mistakes.
problcnr insights ctc.
o Notes and interpretations can start as soon as the field study commences
and till the completion ofthe study.
Research diary
o Keep diaries updatcd, as a docurncntation procedure
o Compare. ifrnore than onc researchcr is involvcd
o Catch intercsting cvcltts in thc diary
o Only docunrcnt that which is essential
o Make rccordings ilrmediatclyaftcran event.

Transcription
When language analysis is involved, the focus in transcription should be to obtain
the maximum exactness in classiffing the statements, pauses, hesitations, etc.

158 SeU:lnst,'tctionol M.deriul

 Table 4. I 2 liwt sc'ript i orr Research Design

t Overlapping speech: the precise point at which one

person begins speaking whilst the other is still talking,
or at whiclr both begin speaking simultaneously, NOTES
resulting in overlapping speech.
(0t2) Pauses: within and between speaker turns, in seconds.
'Aw:: Extend sounds: sound stretches shown by colons, in
proportion to the length of the stretch.
Word: tJnderlining shows stress or emphasis.
'fishi-': A hyphen indicates that a word/sound is broken off.
'.hhhh': Audible intakes of breath are transcribed as '.hhhh'
(the number of h's is proportional to the length of the
breath).
WORD: Increase in amplitude is shown by capital letters.
(words...) Parentheses bound uncertain transcription, including
the transcriber's 'best guess'.

Source: Flick, An lntrodtrclion to Qttanlitalive Reseurch,1998.

Texts produced in this way are the nearly constmcted realities

o The research should talk less and listen as much as possible
o Produce notes as exactly as possible
o Bring data to uniry as early as possible
o The readcrs of the notes must be able to see for themselves clearly
which persons are providing enough data
o The reports should be open arrd clear
o The reseal'cher should seek feedback on furdings/representations
. Presentations should be balanced between the various aspects observed/
recorded
o Writing should be as accurate, as possible
l'hese steps seek to improve validity by sensitizing the researcher to
qualitative research.
Procedural validity is sought to be achieved by the different relationships at
work in rcsearch
o The relationship bet'ween what is observed (behaviours, rituals, meanings,
etc.) and the lager cultural, historical and organizational contexts within
which the observations are made (the substance)
o The relationship among the observed, the observer and the setting (the
context)
o The issue of perspective (or point of view), of the observers' or the
members', used to render an interpretation ofthe ethnographic data
(the interpretation);
o The role ofthe reader in the finalproduct (the audience);
SeU-lnstructional Material 159
Regearch Design o The issue ofrelationsh[, rhetorical, or authorial style used bythe autho($
to render the descripion and / or interpretation (the style).
Here validation involves the entire researchprocess

NOTES Triangulation
This is a term used for cornbining the methods in qualitative research. Four types
oftriangulations are zuggested:
o Data triangulation: It invohes using different data sources in rating persons,
places and situations
o lnvestigatortriangulation: Here different interviewers orobservers are used
with a view to minimizing errors/biases.
o Theory triangulation: It involves approaching data with multfule pospectives
and hypotheses in mind. Various types of orientations are placed side by
side to see their usefulness forproducing knowledge
o Methodological triangulation: It involves combining different methods such
as combing questionnaire with an interview or using different sub-scales for
measuring a phenomenon.
Tiiangulationis seenas aconcept forvalidatingresuhs obtainedwithindividual
methods. These are thought to enrich and complete knowledge and lessen the
limitations ofindividual methods used singly. These are the ways social realities are
sought to be studied systematically. Triangulation is seen as a means to increase
the scope, depth and consistency ofknowledge tlrough methodological means.
Analytic induction: Here the attempt is to understand and explain the
exception that is deviant to a hypothesis in a systematic wayto interpret results. It
is a case of looking at negative data to be able to substantiate the general.

New criteria to evaluate quattative research:

1. Trustworthiness
2. Credtuiliry
3. Dependability
4. Transferability
5. Conforrnability
Credibility can be increased bypersistent observatioq and triangulation of
methods, researchers and data.
Peer meetings is another process
Conrnunication validation
The raw dataare subjected to audits in order to increase dependability
Democratization of qualitative research is achieved by using constant
comparative methods for interpreting texts. This is carried out by
1. Comparing incidents ofeach category
2. Integrating categories, by properties/time, etc.
f60 Self-lnstructional Mateial
,

3. Delimitingthetheory Research Design

4. Writingthetheory
This is a continuous growthprocess. Contrasting cases and idealffi analysis
are carried out so that pure cases can be tracked and the understanding ofthe
individual case be made more systematic. NOTES

Comparisons of old and new criteria for the qualitative field:

ord New
Objective Conformability
Reliability Dependab iliry / Auditability
Internal validity Credibility / authenticiry
Extemalvalidity Transferability / Fittingness
U tilization I app lication Action / Orientation

4.4.14 Presenting the Findings/ Resurts Fo[owing the eualitative

Research
These are to be seen as 'tales from the field' (Flick, 199g)

Realist tales, confessional tales and impressionist tales (dramatic recall)

All these are aimed at providing knowledge
Other forms ofwriting are critical stories and formal stories.
V/ritings never exist in pure form and are never communicable in this form.
The readers have to be kept in mind-who need to have a clear understanding of
what is presented. What the readers need to know in order not to misunderstand
the writing. Therefore, not only how something is presented, but how that is
presented is important. The version, interpretation, and content ofthe text are
inportant fictors in terms ofwriting the findings / results. Here is a diagram ofhow
a text is to be presented:

Construction

Ways to the text:

theoretical sampling,
Constant
.- -- -- companson
: Contrastins
I cases

Communicative
validation

Experience Interpretation

Fig 4.11 Grounding the kxt

Source: Flick, An Introduction to Quantitative Research,lggg.

S e A-In struc tio na I Material 161

Researc'h Design 4,4.15 Computers in Qualitative Research

Computers are widely used in the qualitative research for data analysis. Special
programs are available for analyzing data, combing qualitative and quantitative
NOTES research possibilities. ard transfonning one typc ofdata into another-nualitativc
to quantitativc and vice vcrsa. Triangulation ofrcscarch can also be done with thc
help ofcomputers.

CupcrYouR PROGRESS

I l. When is multiple-regression analysis useful?

12. Give one purpose of texts in a qualitative research.
13. Define the theoryofqualitative research.
14. What does visual data constitute?
15. Definecontent analysis.
16. What are the steps in documenting data in a qualitative research?

4.5 SUMMARY

o A research design is a conceptualfi'amervork for conducting research. It is

a blueprint for collecting, measurin g and analyzing data. Depending on the
topic, time and cost. differcnt types of research designs are used.
o A good rcscarch desigrr should seekto rninirnize bias and maximize reliability
ofthe data obtaincd. Givcn the least possible experimental error, it should
be as objective as possible. No single design should be used for all types of
research problems.
o Quantitative units or values arc quantifiable factors like height, distance,
loudness. etc. Qualitative units are attributes, e.g., honest, extroverte{ shy,
etc. Expcrimental research usually involves quantitative factors, whereas
qualitative research deals with qualitativc factors.
o Qu,alitativc research is highly helplhl in gaining insights into mental disordcrs
and their pattems. Epidcmiological studies show the relationship between
thc frequcncy of schizophrcnia ancl the distribution of this disorder in a
population. Schizophrenia is highcr in the lower socio-economic classes.
These correlations have been well-established by empirical quantitative
data.
o Different types of interviews are used for different kinds of qualitative
rescarches. Some well-known tlpes of interviews include focused interview,
problem-centered interview, expert interview, ethnographic interview,
narrative interview and episodic interview.
162 Sell'-lnshttctional Mats iul
 o observing thc responses of a small group at one time is called a group Research Design
discussion. In a focus group, people are asked about their perceptions,
opinions, beliefs and attitudes towards a product, service, concept,
advertisement, idea, or packaging.
o once data is collected it has to be categorized, codcd and tabulated for NOTES
analysis.

4.6 KEY TERMS

o ANOVA: In statistics, Acollection ofstatistical models and their associated
proccdurcs.
o l\lultiple ANOVA: This is an extcnsion ofa two-wayANOVA.
o canonical analysis: A uscfulmcthod when both the measurable and non-
mcaSurablc variablcs arc prescnt.
o Infercntial analysis: Analysis concemcd with various tests ofsignificance
Ibr hypothcsis tcsting.
o Analytic induction: An attcnrpt to undcrstand and cxplain the exception
that is deviant to a hypothcsis

4.7 ANSWERS TO 'CHECK YOUR PROGRESS'

I . A dcsign is a plan for identifting the sourccs for obtaining information.

2. A variablc with <Iiffercnt quantitative valucs like height, loudness,
is a unit
ctc.and qualitative traits like dcpendable, shy etc.
3. Difhrcnt rcsearch designs include exploratory dcscriptive, diagnostic and
hypothcsis testing designs.
4. Randornization protccts against extrancous factors creeping into a study.
Thc rolc ofchanco factors is also ncutralized by randomization.
5. Matchcd group dcsign is also knorvn as a randomizcd-brock design.
6. Thcrc arc two typcs ofnratching: (i) Matchin-u by pairs--on thc basis of
scorL.s obtaincd. subjccts arc matchcd and paircd: and (ii) Matching by
rne'an and standard dcviation.

7. Wlrcn nrore than tu'o indcpendcnt variablcs arc nranipulated simultaneousty,

thc factorialdesign is used.
ll. A truc cxperimental design means: (i) Equivarent groups are formal,
(ii) Randomization occurs. and (iii) Relcvant factors arc controlled.
9. In a corrclationaldesign, two scts of data are collected from the same
group ofsubjccts so that relationship betwcen two subsequent sets ofdata
can be determined.
Sell'-lnstructional ltlateriul 163
Research Design 10. In a cohort-design, longitudinal studies of several groups are done. It is
useful for gerontological and development studies.
I I . Multiple-regression analysis is a useful method whenthe dependent variable
is influenced by two or more independent variables.
NOTES
1 2. One purpose oftexts in qualitative research is: tests constitute essential data
and are also the instruments forinterpreting data.

I 3. The theories ofqualitative research are circular, not linear. Theories undergo
revisions with new versions ofconstructions and reconstructions.

14. Visual data constitutes observations, photos and fikns.

1 5. Content analysis is a classical method for studying any textual material. It
includes verbal and visual data.

16. The steps in documenting observational data in qualitative research are: (i)
recording the data, (ii) editing the data, and (iii) constructing a new reality
fromthe data.

4.8 QUESTIONS AND EXERCISES

Short-Answer Questions
1. Give anytwo basic principles ofexperimental designs.
2. Give one difference between group and within-group designs.

3. State bne concept related to research design.

4. Why are research designs needed?

5. Why is a research plan developed?

Long-Answer Questions
1 . Mention and explain different research designs.

2. Describe an experimental design.

3. Describe non-experimental designs and their usefulness in research.

4. Describe correlational designs. What is discriminant function analpis?

5. Define qualitativeresearch. What are its aims?

4.9 FURTHER READING

Flick, U. 1998. An Introduction to Qualitative Research. New Delhi: Sage

Publications.

Garfrmkel, H.1967 . Studies in Ethnomethodology. Englewosd Qlift, New Jersey:

Prentice-Hall.

164 Self-Instructional Material

Glaser, B. and Strauss, A.L. 1967 . The Discovery
of Grounded Theory: strategies Research Design

for Qualitative Researcft . New york: Aldine.

Goodman, N. 1978. Ways of World Making. Indianapolis: Hackett.
Kerlinger, F.N. 1973. Foundations of Behavioral Researcft. Newyork: Holt
NOTES
Reinhart and Winstoq Inc.
Kirk, J.L. and Miller, M. 1986. Reality and validity in eualitative Research.
Beverley Hills, CA: Sage.
Kothari, c.R. 1995. Research Methodologt; Methods and Tbchniqares, New
Delhi: Wishwa Prakashan.
Silverman, D. 1993. Interpreting Qualitative Data: Methodsfor Analyzing
Talk, Text and Interaction. London: Sage.
singh,A.K. 2008. Tbsts, Measurements and Research Methods in Behavioural
Sciences. New Delhi: Bharati Bhawan.
Toulmin, s. 1990. cosmopolis: TheHiddenAgendoof Modernifl. Newyork:
Free Press.

Self-Instructional Material 165

 Role of Computers in
UNIT 5 ROLE OF COMPUTERS IN Research, Interprelotion
and Report W.iting

RESEARCH.
INTERPRETATION AND NOTES

REPORT WRITING
Strucfure
5.0 Introduction
5.1 Unit Objectives
5,2 ComputerApplications
5.3 The Computer Systenr
5.3. I
hnportant Charactcristics
5.4 The Binary Number Systern
5,4.1 Decimal to BinaryConversion
5.4.2 Binary to Decilnal Conversion
5,4.3 Computations irr Binary System
5.5 Computers and Researchers
5.6 Interpretation and Reporl Writing
5,6.1 Meaning of Interpretation
5.6.2 The Need and Importance of Interpreting the Findings
5.6.3 Techniques of Interpretation
5.6.4 Precautionsduring Interpretation
5.6.5 Steps in Report Writing
5.6.6 The Layout of the Report
5.6.7 Types of Reports
5.6.8 Oral Presentation of Reports
5.7 Summary
5.8 Key Terms
5.9 Answers to 'Check Your Progress'
5.10 Questions and llxc'rcises
5.ll Further Reading

5.0 INTRODUCTION

In this unit you will learn how computers have revolutionized research work. The
high speed electronic digital computer has a major impact on every phase of
behavioural research. Problem-solving and lengthy statisticaland mathematical
calculations, done manually, are things ofthe past. what took days, weeks and
even months earlier, is now done in a matter ofminutes. Research studies and
calculations, which looked impossible earlier can now be tackled with the aid of
computers, in minutes or hours. Computers are now dominating almost every
walk oflife, which not only makes them important but indispensable. Here, we will
look at the applications ofthe computer, some important characteristics, the binary
number systernand the role ofcomputers in research.

Self-lnsttttttiotrul Mqteriol 167

Role of Computers in
Re s e arch, I n terp r e t q t i o n 5.1 UNIT OBJECTIVES
and Report Witing

After going through this unit, you will be able to:

NOTES o Understand the essential characteristics of a computer.
o Understand the binary number system and learn important conversions
o Realize the importance of computers in research and learn how to use them
o Understand and interpret report writing andthe relevant techniques
o Know about the different types ofreports and their features

5,2 COMPUTER APPLICATIOI\S

Today, computers are used in allpossible fields and forvarious pu{poses. Every
sector, be iteducation, cofirmerce, nunagement, industryorcommunications, relies
on conputers for its snrooth functioning. Even ifan individual is not directly involved
with the functioning of computers, his / her life is affected by them in everyday life
andwork.
Computers are not only used in numeric applications like carrying out
complex research and data ana$sis, but also for non-numeric uses like assisting in
teaching and learningprocesses, providing a large databank ofinformation, handling
payrolls, record keeping, financial forecasting, making clinical diagnosis, providing
entertainment like playing games, watching movies and listening to music, besides
sportsviewing.
Computers are used in applications ranging from running a firmto monitoring
all environmental effects. Computers have made the research and the development
ofvarious diagnosis and prevention methods, easy and cost-effective. For exanple,
a computer can accuratelyprovide better forecasts when andwhere an earthquake
or tsunami is likely to occur, the effects of drugs on the human systenr, dissections
for study purposes, creating three-dimensional models for buildings or airplanes,
and a host of other applications in media that have brought about a revolution in
communications.

5.3 THE COMPUTER SYSTEM

The computer is aprogrammable machine that receives data, analyses it and gives
a useful output. Aprogram is a set of instructions given to the computer for a
parlicular kind offi.rnctioning. New and better progftrms are being written everyday
for easier and greater uses. The computer system includes not only the hardware
with an operating system, but also the software that is necessary to make the
computer function. Complex programs are written for handling a varietyoftasks.
Every computer systemrequires an operating systern An operating system
is a software program that enables the computer hardware to communicate and
16E Self-lnstructional Material
 operate with the computer software. The following diagram illustrates
how the Role of Computers in
computer systemworks: Research, Inlerpretation
and Report Writing

F.r
rt NOTES

fi
iL

ru
0
l;;r
t"-:_]
Fig. 5.1 Coinputer Architecture

cPU: The centralprocessingunit or cpu is responsible forrunningprograms

and applications on the computer. The cpu has three segments: l. Intemal.to*g.,
2. Control unit, and 3. Arithmetic logic unit

A conputer program is wriuen into the intemal storage and then transmitted
to the control unit. This data is available for processing Uy ttre arithmetic logical
unit, which conveys the results back to the internal storage, and thus the
ouput is
obtained fromthe intemal stage ofthe CpU.
The four primary components of a computer system are:
Input: Input devices senddata and instructions to the CPU. Theytranslate
the datainto binarylanguage whichthe CpU understands.

output: output devices makeprocessed data available to the user. The

output devices reiterate the binary characters into familiar language characters.
Processing: cPU executes instructions supplied by the program.
secondary stora ge : Primary memory temporarily ho lds limited amounts
of data and programs whereas secondary storage devices permanently
store data
and progranl- of great amounts.

The four essentialhardware components ofa computer system are:

input,
processing, output and storage systems. physical components
ofcofiputer rystern,
including the electronic and mechanical parts, are called computer hardware.
The
intangible components like data and programs are called computer software.

Self-Inst,actionol Material 169

Role of Computers in Software components are stored on hardware components like cache in CPU,
Research, Interpretation
and Report Writing
hard disks, CDs, DVDs, tape, pen drives.

5.3.1 Important Characteristics

NOTES o Speed: The computer is a very high speed electronic device. Computer
can performmultiple operations on data in seconds, depending onthe speed
with which it operates. Their speed is usually measured in megahertz M{z).
Different computers have different speeds.
o Diligence: Being amachine, the computer does not suffer fromfatigue,
boredom or lack ofconcentration. It can work continuous$ for hours without
creating any elror. It performs its operations with the same accuracy and
speed throughout.
o Storage: A computer's secondary (external) storage can hold large amounts
of data and programs which can be retrieved as and when required can be
stored for firther use. The stored data is available anytime for processing.
A cache is a storage device inside the CPU (internal) whichcan store small
amounts of essential data for temporary use.
o Accuracy: The computer's accuracy totally
is consistently high. Output is
dependent on the given instructions and input data. Ifthe data provided is
right, the output will be accurate. In the event an elror occurs, the error
detecting techniques are so efficient that it seldom leads to false results.
o Automation: A computer can perform automatically once the program is
in the computer's memory. It controls different devices attached to it and
executes the programinstructions one byone.
o Binary digits: All the instructions provided and data stored in computers
are in the binary number format which consists only two digits I and 0. This
makes everything simple, accurate and fast.
o Arithmetical and logical operations: Arithmetic operations include
performing additioq subtraction, muhfulication and division onthe numerical
data. Logical operations compare the numerical and also alphabetical data'
o Reliability: The extremelylow failure rate ofmodern day conputers has
made their results consistent. In general they are reliable.
o Retrieving data and programs: The stored data and programs can be
retrieved very quickly for further processing.
o Versatility: Computers can perform different kinds of tasks. One side
you canplay games, on other side you can send emails, etc'
. Communications: Today conrputers ale emerging as most powerful mean
of communications. You can communicate via computers using emails via
LAN, WAN andlntemet. Youcanalso talkthroughlnternetphone like
the
GoogleTalk.

Se lf-I nstru c t ion a I Ma teri a I

 Role of Computers in
I n t erpr e t at i o n
Re s ear ch,
CmcxYoun Pnocnnss and Report Witing

l. List the various uses ofa computer.

2. Give an example ofbinaryto decimal conversion. NOTES
3. Give an example ofdecimal to binary conversion.
4. Give one important characteristic ofa computer?
5. What is the frrnction ofCPU?

5.4 THE BINARY NUMBER SYSTEM

Binary means having two enitities. In binary mathematics system there are only
two digits 0 and 1. Inthis sytem, allthe numbers, alphabets, instructions, etc are
expressed in terms of0s and I s. on the other hand, in decimal system there are
ten possible digits, 0 to 9. Since computers use only binary systerl the operations
are simple, quicker and accurate.

we all know that computers can calculate complex equations and perform
complex mathematics at lightning speed. Although a computer will only
process ls and 0s, there comes apoint when the ls and 0s have to be converted
into the usual decimal nurnbers that we all are familiar with.
Let us consider a number 1234:
Thousands Hundreds Tens Ones
r234
Whichmeans,
1234:1 x 1000+2 x 100+3 x l0 +4x I
The binary system operates with base 2 or radtx2 (6i is the Latin prefix for
two) or it uses0s and ls to represent numbers. The simple comparison between
decimal and binary are given in the following table:

Thousands Hundreds Tens Ones

Decirnal I 0^3 l0^2 l0^l l0^0
Binary 2^3 2^2 2nI 2^0
In base 1 0, we put the digits 0-9 in columns I 0^0, I 0^ I , I 0^2, and so on.
To put a nurnber that is greater than 9 into l0^n, we must add to l0^(n+ I For
).
example, adding 1 0 to column I 0^0 requires us to add 1 to the column 1 0^ 1 .
Inbase 2, we put the digits 0-1 in columns 2 0,2^1,2^3, andso on. To
put a ntrnber that is greater than I into 2^r1 we must add to 2^(n+ I For example,
).
adding 3 to column 2^0 requires us to add 1 to the column 2^1.

Se lf-In s tmc tio nal Material t7t

Role of Computers in 5.4.1 Decimal to Binary Conversion
Research, Interpretation
and Report Witing
A positive decimal integer can be converted into an equivalent binary form by
dividing the decimal numhr repeatedly by 2.
NOTES Start by dividing the given decimal integer by 2.
Let r1 be the remainder andql the quotient.
Next, divide grby2 and let r, and Q, be the remainder and quotient
reqpectively.
Continue this process ofdivisionby2 until a 0 is obtained as quotient.
The equivalent binary number can be formed by arranging the remainders
as rk ..ro-,. . ..r,
(rk and r1 are the last and the first remainders respectively, obtained by the
division process.)
Illustration 5.1 : Finding the binary equivalents of 26 and 45.

Solution:

Number to be Quotient Remainder

divided by 2
26 13 0
13 6 I
6 J 0
a
J 1 I
1 0 1

26 (decimal) : 11010 (binary)

Similarly, we c:;n find the binaryequivalent of45 as illustrated below:

Number to be Quotient Remainder

divided by 2
45 22 1

22 11 0
11 5 I
5 2 I
2 I 0
I 0 I

Thus the binary equivalent of 45 is 10 I I 0l

172 Self-lnsrruclional Matetial

 Here are some decimalnumbers represented inbinary. Role of Computers in
Research, Interpretation
and Report 14?iting
Decimal Binary
I I
2 l0 NOTES
J 11
4 00
5 01
6 l0
7 1l
8 1000
9 l00l
10 l0l0
5.4.2 Binary to Decimal Conversion
This particularmethod is knownas the double-babble method.
To convert a binary number into decimal number, double the leftmost bit of
the given binary number and add to it the bit to its right.
Then, double the sum and add to it the third bit from the left.
Proceed in this manner till all the bits have been considered.
The final sum obtained by repeated doubling and adding is the desired
decimalequivalent.
Illustration 5.2: Converting the binary number 1 I 0 I into its decimal equivalent
using the method given above.

Solution:
l. Doubling the leftmostbit we get 2
2. Adding the bit on its right 1, we get 2-tl :3
3. Doubling again this number obtained (3), we get 6
4. Adding to it the next bit 0, we get 6+ 0: 6
5. Againdoublingwe get 12
6. Finallyadding the last bit (1), we get 12+ I = 13
The decimalequivalent ofbinary ll0l is 13.
5.4.3 Computations in Binary System
Binary addition: binary addition is just like decirnal addition. It follows as:
0*0:0
0* 1:1
l+6:1
l+1:1g
Self-Instructional Material 173
Role of Computers in Sum of 1 and 1 is written as '10' (0 as sum and 1 as carry), which is the
Re s ear ch, I n terp re tat i o n
and Report Witing equivalent ofdecimal digit'2' .

Illustration 5.3: Adding the binarynumbers 1010 and 101.

NOTES Solution: Add 1010 and 101

Binary Decimalequivalent
1010 (10)
+ 101 + (s)
:1111 15

The computer performs all the other arithmetic operations, namely

muhfulication, subtraction, etc. bytreating it as repeated addition a form ofaddition
For example, the multiplication 6*8 is evaluated by repeated addition with the
necessary carryovers, 8+8+8+8+8+8. Though this process seems to be a longer
method, the computer is well-suited to carry out these operations at great speed.

Converting Binary Fractions into Decimal Fractions

In decimal systenq we use a decimalpoint to separate the whole number and the
decimal fraction parts of a given decimal number. In the same way, we can use a
binarypoint to separate the whole and fractionalparts inbinarynumbers. The
binaryfractioncanbe converted into decimal fractions as showninlllustration4:
Illustration 5.4: Converting the binaryfraction 0.101 into its equivalent decimal
form
Solution:
0.101 (binary) : (l) Il2+ lqr14 + (1)1/8
: 0.5 + 0.0 + 0.125
: 0.625 (decimal)
To convert the decimal fraction to binary fraction, the following rules are applied:

1. Multiply the decimal fraction repeatedly by 2.The whole number part of

the first 1 or 0 ofthe binary fraction;
2. The fractionalpart ofthe result is carried over and multiplied by 2;

3. The whole number part ofthe result gives the second 1 or 0 and so on.

Illustration 5.5: Convert 0.625 into its equivalent burary fraction.

Solution: Let us convert 0.625 into its equivalent binary fraction.
0.625*2: 1.250 - I:0.250
0.250*2 : 0.500 -0 : 0.500
0.500*2 : 1.000-1 :000
Hence, 0. 101 is the required binary equivalent.

174 SelfJnstructional Material

 We can now see how computer arithmetic is based on addition. While Role of Computers in
Research, Interpretation
performing such calculations, the number ofindividual steps increases enormously and Report Witing
because all computer arithmetic is reduced to addition. However, since the
computer is capable ofcarrying out millions ofbinary additions per second, it will
not be a disadvantage. NOTES

5.5 COMPUTERS AND RESEARCHERS

The computer has emerged as one ofthe most useful research tools in modern
times. It does a great variety ofjobs with tremendous speed and efficiency.
Computer usage has become a subject of study at all the schools and today
computer is an indispensable part ofany profession.
Conputers have become high$ usefirl tools in the research process, particular
whenthe datais largeandconplex involving corrplicatedrnathematicsandstatistics.
Researchers often deal with huge amounts of data, needing timely storage and
quickretrievalwhenrequired and dataprocessing usingvarious techniques. Apart
from speeding the research process, computers have also added quality to the
researchprocess. Computers assist the researchers throughout the different phases
ofthe research process.
Phases ofresearch process :

L Conceptualphase
2. Design andplanning phase
3. Empiricalphase
4. Analytic phase

5. Disseminationphase
6. Concludingphase
1. Role of computers in conceptual phase: Conceptual phase is when the
researcher formulates the research problenr, reviews literature and formulates
the hypothesis.

Computers help in searching for the literature through the Internet from
databases stored in servers all over the world. The desired information can
then be downloaded and stored in the computer for future use. It thus saves
the time spent on visiting libraries and collecting the data by writing down
the relevant material.

2. Role of computers in design and planning phase: Designandplanning

populatioq identifying research
phase consist ofresearch desigrr, defining the
variables, determining sanpling plaq reviewing researchplan and performing
pilot study.
Several software programs like NCSS-PASS-GESS are available to
calculate the sample size required for a proposed study. The standard
Self-Instmctional Material 175
Role of Computers in
deviation ofthe data fromthe pilot study is required for the sample size
Res earch, Interpretat ion
and Report Wliting calculation
3. RoIe of computers in empirical phase: This phase consists ofcollecting
and preparing the data for analysis" The data obtained is stored in computers.
NOTES
This can be edited, represented graphically thereby helping the researcher
with data entry data editing and data management.
4. Role of computers in data analysis: This phase consists of statistical
analysis ofthe data and interpretation ofthe results. Manyprograms are
available to perform advanced calculations using statistical methods. They
can be used to calculate the sample size, test the hypothesis and calc:.1ate.
Compri-ten not onlyanalyse but also monitor the acctracy and completeness
ofthe data. Computation ofthe 'mean,' 'standard deviation,' 'correlation
coefficients,' factor ana$sis and nultiple non-parametric analysis are just a
few ofthe statistical calculations that the cornputers can performwith ease
and accuracy, besides speed.

A note of caution
It computer is just a tool and a resource. Computers
is wise to remember that a
can only compute but cannot think. Computer-based analysis is not usually
economical in case of small research projects. Certain important details which are
saved in the computer may get lost.

conputer functions onthe basis ofthe programwritten. Adequate planning

and suitable organization is necessary for the efficient handling ofdata.
To sumup, computers make the research process easier, faster, accurate,
reliable and less tedious. Howeveq the experimenter should have thorough
knowledge about the abilities and the limitations ofthe software used for the better
use ofcomputers.

The use ofcomputers enables the researcher to use trial and effor processes
which involve a lot ofcalculations andrepetitive work. Not onlydoes it produce
the results rapidly but the different options are also made available to researchers.
These would otherwise take days or months.
Researchers must be fimiliar with data organization and coding, storing the
data in the computer, selection of appropriate statistical techniques, selection of
appropriate software packages and execution ofthe computer progralr

5.6 INTERPRETATION AND REPORT WRITING

Once the studyhas been completed, the analysis is carried out and the conclusions
drawn, the report ofthe entire studyand the findings have to be written. It is only
through the process of interpretation that the researcher can tell the scientific
community about the relationships that were studied and the theoretical concepts
underlying the findings.

176 Self-Instmctional Material

 The types ofanalysis that the data was subjected to and the inferences that Role of Computers in
were drawn also need to be readily communicated in the report writing. Other Research, Intelpretation
and Report Witing
researchers look forward to such an endeavour from the investigator.
The meaning of writing a report, the importance of interpretation, the
techniques ofinterpretation and the cautions that need to be exercised are presented NOTES
in this section. The significance ofscientific writing as well as the steps involved in
reportwriting is suggested. This is followedbythe layoutplanofaresearchreport.
5.6.1 Meaning of Interpretation

The purpose ofwriting aresearchreport isto tellthe scientific communitywhat the

studywas, how it was done, the results obtained and the conclusions drawn. All
interpretations have two major fiurctions: 1. To establish the findings ofthe current
study and to show its links with other prior studies in the related field. This is to
help in the continuity of research; and 2. To provide explanations in terms of
relationships studied as suggested by the data obtained from the study. These twin
attempts have the goal ofadvancing knowledge and stimulating fi.yther research.

5.6.2 The Need And Importance Of Interpreting The Findings

Interpretationprovides satisfactoryexplanationto the findings andthe usefulness
a
ofthese for any research. It is through this process that the abstract principles
underlying any findings can be clearly understood. Interpretation serves as a
framework for linking the current studywith other studies inthe same field.
Interpretations provide the inrpetus for further future research. It is also a
means for stimulating new avenues in the quest for knowledge. Interpretations
also throw out the real significance ofthe findings ofthe study. In exploratory
studies interpretations become the basis forhypothesis formulation forexperimental
research. These are termed post-factum interpretations.
All these indicate the need for interpretation of findings following research
studies.

Interpretationmeans drawing inferences fromtheresearchdata. It establishes

continuity in research studies. It acts as a guide for further research sfudies and
explains the concepts that have been observed. It is a skill that can be best learned
through practice and experience. Researcher must explain his concepts with the
help of theory and keep extraneous information and all relevant factors in
consideration while interpreting research studies. The researcher must try and
make sure that the source of data is trustworthy and accurate, and that the data
reflects good homogenerty. It is important to use proper statistical methods for
analysis. It is important to be aware and take precautions against all possible
sources oferror. The researcher also has to keep a check on his tendency to
over-generalue or be biased about the results. Also there should be constant
interaction between initial hypothesis, empirical observation and theoretical
conceptions.

Self-Instructional Material 177

Role of Computers in There are certain steps that a researcher should follow while writing a report.
Res e ar ch, I nte rpr e t at i o n
and Report Witing Firstly, the material should be analysed either logically or cluonologically. After
this, an outline should be prepared ofthe contents that are to be included in the
report. Afterthis, aroughdraft containing the procedure adoptedto collect data,
NOTES technique ofanalysis, findings and generalizations is put down. This is then revised,
polished and rewritten to make it more clear and comprehensive for the reader.
After this step, the bibliography is prepared in an abhabetical manner.
Basicallythe report canbe divided into three broad categories.
(i) Preliminary pages: This section carries the title, date,
acknowledgements, preface, contents, and a list of tables and
illustrations used throughout the report.
(ii) Main text: The main textincludes the introduction, statement of
findings andrecommendations, results, implications ofthe result and
sunmary.
(iii) End maffer: Here the appendices me listed along with the bfoliography
ofall sources consulted or relevant to the research study.
Research reports vary in length and format. Research reports can be in the
form of a technical report, popular report, an article, a manuscript or even an oral
presentation.
A technicalreport is a detailed report required for record keeping or public
dissemination. It details the methods used, assumptions made during the course of
the study and focuses on detailed presentations of the findings including their
limitations and supporting data. It should be substantiated with charts and graphs
so that it is clear and easilyunderstood.

A popular report is often simple and attractive. It does not use too many
mathematical or technical terms; instead it uses charts and diagrams, large fonts,
manysubheadings andvisuals whichnrake it easyto read. It enrphasizes onpractical
findings and their implications.
Along with the oral presentation, there should also be a written document
for record keeping and can also be circulated before the presentation to acquaint
the audience. It should be supplemented with visuals, slides, wall charts and
blackboards. Abrief outline should also be given.
All these fictors highlight the need for interpretation offindings ofthe research
studies.

5.6.3 Techniques of Interpretation

The task of interpreting the findings of a study are demanding and involve
considerable skills acquiredthroughexposure andpractice. It is appropriate to
seek expert advice as and when needed to interpret one's findings. The techniques
involved in the process ofinterpretation have several steps as list below:
o All relationships that have been found in the course ofthe study must be
explained adequately. Any common link that ties the findings together must
178 Se lf- I nstru cIio na I Ma te ri a I
 be identified, clearly. This is the beginning ofthe process ofgeneralization, Role of Computers in
Res ear ch, I n t erpre I al i o n
following a study;
and Report Witing
o Data tl:rit have been thrown up by the study incidentally or accidentally also
needs to be given sufficient importance while interpreting the findings that
one was studying. These bits of information may have far reaching NOTES
consequences for the current study, as well as those in the future.
o one can always open the studyto an expert or a practitioner in the given
field to act as the devil's advocate, so that any omissions or fallacies in the
logic of interpretation can be detected. This would enable correction
opportunities.
o All facts must be obtained and collated in the most suitable manner before
interpretations are carried out. This would rule out the chances ofhasty or
premature conclusions.
These techniques are suggested to ensure truly valid and generalizable
conchrsions from the study.

5.6.4 Precautions during Interpretation

Merely carrying out a well-designed study and obtaining properly collected data,
does not preclude the possibility oferroneous interpretations. To prevent drawing
ofwrong conclusions; a few precautionary steps are indicated and suggestions for
correct interpretations are given below:
Some ofthe precautions to be exercised are:
o Ensure that the data arcreliable and valid;
o Ascertain the homogeneity ofthe data, collected.
a Make sure that the data have been subjected to proper statistical analysis.
o Double checkto see ifthe interpretation ofthe statisticalsignificance of
data is well within the prescribed range.
o It is important to note that correlational data are not only statistically
significant, but also interms ofthe meaningfulness ofthe studyresults.
o where causaleffects are determined, it is absolutelynecessaryto ensure
that causation has indeed occurred. These are critical for avoiding false
conclusions and generalizations.
o A second round ofcomputational checks, verification ofsignificance need
to be carried out to prevent effors from creeping in.
o Restrict the generalizations and interpretations to the study undertaken and
never over-extend. Limiting the scope ofthe interpretations ofone's frrdings
is asafe optionto prevent inaccurate conclusions.
o All interpretations must ultimately go back to the starting point oftlre study
in terms ofthe research problemthat is being investigated. Findings must
always be made keeping with this connection in mind. The theoretical
concepts, the empirical observations and the analysis carried out on the
Self-Ins*uctionol Material 179
Role of Computers in data obtained should ideally be in synctrony with the findings and
Research, Interpretation
and Report Witing interpretations o ffered.
These alone would ensure effor free interpretations

NOTES 5.6.5 Steps In Report Writing

Since research is a demanding andpainstaking effort, sufficient time and attention
should be devoted to writing the report ofthe work carried out. Some ofthe steps
invohedintheplanning ofwriting agood report are indicatedbelow forthebenefit
ofresearchers:
o The entire study should be analyzed very logically in a thorough manner.
o A rough-or preliminary draft ofthe outline ofthe proposed research, should
be prepared withthought and care;
o Then the final outline should be developed.
o This can then be polished and refined, before being re-written.
. This canbe followed bythe compiling ofthe bibliography.
o Then the final draft can be written.

The logical analysis process

This can be carried out in one oftwo ways:
(i) Advance the subject ofthe study logically. This means one topic follows the
otheq rationally.
(ir) Chronological This invohes writing a report onthe basis ofthe sequence of
happenings.
Sometimes, a combination ofthese two strategies is useful.
Preparation of the rough-draft: This is the step ofputting down what
one did in the study, in the order in which it was carried out. This could include the
motivations, the insights, the struggles andthe limitations. Also, theprocess of
designing the study, the method ofdatacollectiorl the controls andvariables should
be utilized and the broad findings and directions ofgeneralizations possible should
be identified.

Preparation of the final outline: This step involves careful reporting of

the study along with the theoretical framework for the study. List all the procedures
followed
The prcparation of the bibliography: This refers to all the books, joumals,
magazines, reports andother sources that have beenconsulted forundertaking
the study. The format for listing these references is to be written in keeping with the
style of 'Manual of the American Psychological Association'.

5.6.6 The Layout of the Report

This is the structure or format of a research report. There is a standardued
convention that has been universally accepted for the purpose ofwriting a research
180 Self-lnstructional Material
report. It is clearly laid down in the publication, 'Manual of the American Role of Computers in
Reseorch, Interpretation
PsychologicalAssociation.'Thismanualalso gives suggestions onhowto present and Report LVfiting
concepts with economy and clarity. The APA manual has divided research report
writing into eight parts.
NOTES
APA format for writing reports
(i) Title page: This page has three main sections:
a. Title and running head. Thetitle is the probleminvestigatedbythe
study while the running head is a short description ofthe Title. It can
have a maximumof50 characters;
b. Author's name and affliation: The autho(s) nanre (s) to be presented
on this page, starting with the principal investigator and the designation
and the institutional affiation (ifany). The other authors ofthe study
should thenbe listed in the order oftheir importance to the contnbution.
Their designations and affiliations can be given suitably.
c. Acknowledgements: This can be third part ofthe title. Any form of
support in terms offinances or institutionalhe$, comments, secretarial
/ editorial assistance etc, can be mentioned here.
(i| Abstract: This is a synopsis ofthe research in full detail, but in an abridged
form. The ideal length would be about 150 words. It should include the
problem studied, the methods used, the research design adopted (any
apparatus, ifused), the statistical treatment, the findings and the conclusions,
as well as its applications / implications.
(iif Introduction: This is the beginning ofthe report and thereafterthe entire
report follows a sequence. An introduction is not written simply as an
introduction. It has three divisions inthe report writing.
a. The fust part is the statement ofthe problem that was studied. Apart
ofthis sectionneeds to include the motivation for studyingthe particular
problem and the theoretical or practical orientations underlying the
study undertaken.
b. The second part ofthis section should contain a description ofthe
review ofthe previous literature avaihble in the field. This should refer
to the studies related to the field ofthe present inquiry. A good rationale
and a logical relationship between the earlier studies and the present
investigation needs to be developed.
c. The third part of this section is devoted to the formulation of the
hypothesis, for the study. This should inchrde the operational definition
ofthe hypothesis that is proposed. The hypothesis is to be clearly
stated and the scientific procedures involved in testing it need to be
specified. The independent and dependent variables should be
identified and the design is shown in a diagram, ifpossible.
(iv) Method: This constitutes the main body of the report to be written. .It
includes a veryprecise account ofhow the study was conducted. This is the
Self-lnslructional Mqterial l8l
Role of Computers in procedural part ofthe report. It is expected that anyone who reads the
Res e arch, I nt e rpr et at io n
and Report Witing report and wishes to undertake a similar study should be able to do so, by
following the method's section closely. There are several sub-sections that
make up this section. These help inbringing to focus different parts ofthe
NOTES method:
a.The sample: This requires that the sample be clearly specified, along
with the type of sanrpling procedure used. Special characteristics like
age, sex, etc., that are relevant to the study needs to be identified
precisely. The total number of subjects included for the studyis very
significant. The different treatments should be clearly described. The
control conditions should be explained carefully.
b. Equipment used: Ifany special apparatus was used for the study, the
model, the manuftcturing conpany and other details need to be given.
If a photograph ofthe equipment can be given, it would be helpful.
c. The design: Every study has a blueprint on the basis ofwhich any
study is proposed to be carried out. This is the design ofthe study.
Here, the entire procedure ofthe conduction ofthe study, along with
the experimental and control conditions, need to be elaborately
presented.
The steps in the introduction ofthe independent variable have to be dealt
with at length and clearly. So, also the controls that have beenused in the
study, the nature and type ofmeasurement ofthe dependent variable should
be clearlymentioned. In short, the entire procedural information has to be
given, so that its replication is possible by another interested researchers.
(v) Results: This refers to providing the data obtained from the study. Ifthe
raw data is significant for being presented, it is to be given in a meaningful
tabulated way. Otherwise, the results are to be reported in terms of the
hypothesis studied. The statistical significance ofthe results needs to be
given. Whether the hypothesis is supported or rejected should be clearly
mentioned. Tables, charts, diagrams, figures, etc. areto bepresentedinthe
most suitablemanner
The major findings can be summarized and presented. Wherever the texts
can be supported by visual material, it can be done, to make for greater
ernphasis. The statisticaltreatnrent carried about and the levels ofsignificance
for the data should be clearly stated.
(vi) Discussion of results: The purpose ofthis section is to interpret the results
ofthe study. It should have a detailed account ofthe study related to other
studies in the field. The discussion should state whether the hlpothesis is
supported or rejected. The supported hypothesis, as well as the rejected
one, both need to be explained on the basis ofsome theoretical assumptions.
Any new hypothesis can be suggested when the results are not along the
predicted lines. Anyfaults in the formulatedhypothesis canbe modified on

lEz Self-Instruclional Material

 the basis ofthe results obtained. These can be discussed suitably. Detailed Role of Computers in
Res earch,I nt erpr e t a t ion
discussion about the findings canbe presented. Finally, any suggestions
and Report Witing
regarding the ways in which the problemthat has been studied, need to be
resolved, canbe offered. Any implications ofthe findings canbe indicated.
Also suggestions for future research should be mentioned. A small paragraph NOTES
could be devoted to the limitations ofthe study and details about overcoming
some ofthese can also be mentioned.
(vii) References: This includes all books, journals, articles, reports, internet
links for references, etc . , have to be listed in alphabetical order. The citations
should be presented in accordance with the manual of the American
Psycho lo gical Association.
(viii) Appendix: This canbe included where tests, questionnaires, anystatistical
treatment, computerprografiL etc, were used forthe studybut do not fit
into the report due to length and other considerations.

Style of writing report

Reports have to be clear, concise, complete and easy to read. The object of
reporting a study is to help others who read it to be able to follow the procedure
and carry out a similar activity. The report should be clear in description and
explanation. The reference works must mention authors by last name, only.
Abbreviations are to be avoided, as a rule. only the verywell known abbneviations
such as IQ canbe used. Detailed formulae and computations canbe eliminated.
The report has to be written in the past tense. All sources must be
acknowledged, correctly and completely. Certain permissible and well known
acronyrns alone are allowed to be used in report writing, e.g., STM (Short Term
Memory). This can also figure inthe midst ofa sentence, but not at the beginning.
Ideas must be paraphrased and written. As a rule, whenever quotes are used, the
relevant studyshould be cited.
As for as possible, use onlythe accepted terminology. Numbers that come
in the report which are from zero to nine have to be written in words, while those
from l0 or larger can be written in digits.
5.6.7 Types of Reports
Reports vary in length, format and purpose. The research report presented here is
one ofa scientific study. Other forms ofreports are:
(i) Business reports are largely in the form of a letter. They are short, precise
and factual in detail.

(i) Banks and other financial institutions present their reports in the form ofa
balance sheet. This is largely a statement of accounts for their customers
and shareholders.

(iif Chemists present their reports in terms of formulae and other symbols to
explain the preparation, etc.

Self-lnstructional Material 183

Role of Computers in (iv) Manufacturing companies prepare reports in the form of manuals of
Research, Interpretation
and Report llTiting operation, servicing, spares, etc.
(v) Students ofliterature, linguistics, philosophy and others write reports that
criticallyana$se a choice subject matter. Here quotations and descriptions
NOTES
arecommon.
(vi) Newspapers, magazines and other media personnel present items or articles
in the form ofreports. These include first hand account ofevents or opinions
on a topic. Sometimes interviews ofexperts in a given field are reported on
a theme. This kind of report is largely fact-based, with some details to
explain the happenings. Some of these reports have e.ve-witness accounts
also. The implications of these events form the later parts of the report
writmg.
(vii) Book-review reports analyse the content, style, format, language and other
details ofa book, so as to inform the readers about the book. This is often
a short report.

(viii) Government reports / reports of commissions are very elaborate reports

that are comprehensive in nature. These reports are the result ofa detailed
inqurry carried out on a topic of interest, nationwide or about a specific
concem / matter.
(ix) Case history reports submitted by doctors and other healthcare personnel
involve in-depth studies ofcases. They are based on diagnostic findings
and include plans of treatment.

Research result presentation styles

There are two ways in which the findings of a research study can be presented:
(i) A technical report and (ii) A popular report.
(i) Technical report: This is a veryprecise and detailed report ofthe problem
studied, the methods used and the results obtained, along with some
limitations. It is 'how it was done' report. The findings of the study are
crucial in such a report. Therefore, it begins by stating the results ofthe
study. The format is as follows.
a. Summary of the results: Astatement ofthe main findings. This can
be in some detail as the readers are eager to know the results.
b. Nature ofthe sntdy: This has the general description
ofthe study, the
operational definitions involved the hypothesis formulated, the data to
be collectedandthe type ofanalysis to follow.
c. Methods: This section includes steps like sampling design, sample
selection sample size and other similar details. Any study design if
applicable.
d. Data: How the data was collected. The sources from where the data
was obtained and other related information are included under this
head.
184 SelfJnstructional Material
 e. Data arcalysis andfindings: The statistical treatments to which the Role of Computers in
Re s e ar ch, Int
e rp r e t at i on
data were subjected to are to be described. The findings are to be
and Report l4?iting
presented in tables, clarets, diagrams etc to be given here, for easy
turderstanding.
f. Conclusiorzs.' The detailed summary ofthe findings and the inplications NOTES
ofthe results, are to be presented.
g. References: The diffbrent sources which are cited and consulted for
the study, to be mentioned in complete, detail.
h. Appendix: Questionnaires, lists, formulae and other such information
to be provided in a handymanner.
The entire report must be precise, readable and simple in style. charts,
diagrarns and other supporting material should be presented wherever
suitable.
(ii) Popular report: The eim of such a report is to familiarue the average
reader with some important scientific studies and their findings. so, the
language should be simple, direct and familiar to the reader. So avoiding
technicaljargon, as far as possible, is suggested here. practical applications
ofthe findings ofthe study r,vill have important place in such a report.

The following is the suggested outline, for such a report

a. Thefindings results and their implications.' Since the focus ofthis
report is on its useflilness; the priority lies in the sub-heading taking
the lead.
b. Findings that can be used: Here the suggestion is how the study
conclusions can be used readily. This is an action-oriented detail.
c. Thepurpose af the study:Thereader canbe orientedto the original
reason for undertaking the study and the specific details involved.
d. Methods used: This should be a non-technical description of the
methods used for the study as rvell as some details about how the
sample was chosenand the data collected.
e. Results: This is ofmain ilrterest to the reader. So the conrplete details
ofthe findings along with the best possible illustration ofthe results in
charts, diagrams, etc., are to be presented here. The language must
be as non-technical as possible.

f. Appendix.' All types oftests, questionnaires. and other details that are
ofinterest to the public is to given here.
Make the report simple, easyto read and understand apart from an appealing
presentatior,.

5.6.8 Oral Fresentation of Reports

Oral presentations are dynamic and interactive wa-vs ofcommunicating research
findings. At times, oral presentations are to be made to those who wish to take

Self-lnstructional Material 185

Role of Computers in decisions on the basis ofthe findings of a study. Policy makers are such audiences.
Research, Interpretation
and Report Witing
They wish to see and hear the results of a study for a quick understanding and
easy decision making. It is always necessary for a written report to be circulated,
before the oral presentation is made.
NOTES
Key to effective oral presentations:
. Have liberaluse ofvisual aids
o Use colour generouslyto highlight features
o Have pointed effective captions and titles with the slides.
o Illustrate with graphics wherever possible.
o Have bold fonts to show conclusions, implications, etc.

How to retain listener/vier.ver attention in an oral presentation:

(i) Begin with an anecdote, experience or an illustration
(if Then present the key aspects ofthe study
(iif Identifz 3 to 5 major points to be rnade during the talk
(iv) Connect ideas, logically and proceed
(v) Give examples as you Proceed
(vi) Make eye-contact withthe audience, always
(vii) Modulate your voice through the talk
(viii) Use body language to good efFect
(ix) Bemindful ofthetime allotted
(x) End on a strong note (leave the audience with a thought or an application of
the results)

(xi) Always, be available for taking questions

(xii) Never forget to add a dash ofhumour in your prescntation
(xiif Always have a speaking outline in front ofyou and the rest can be on the
projection screen.

5.7 SUMMARY

o Computer is a programmable machine that receives input, stores and

manipulates data and provides output in a useful format.
o Conrputers are not only a necessitybut theyhave become an indispensable
part o f research process.
o The four primary components of a computer system are: (i) Input devices,
e.g., keypad, which send data to the central processing unit (CPU); (ii)
Output devices, like monitor, which provide the processed information (or
results); (iii) Processor, e.g., CPU, whichexecutes computer instructions;

186 SeU:lnstructional Mqterial

 and4. Secondary storage like hard disk, CD, DVD andpen-drive where Role of Computers in
Research, Interprelation
the information is stored for further use.
and Report Writing
o Conputers are characterizedbyhigh speed, diligence, accuracy, automatiorl
storage capability, reliability and versatility.
NOTES
o Computer calculations are based on the binary system unlike the decimal
system that we use. The binary system uses only two digits, ls and 0s.
Decimal numbers can be converted to binary and vice versa. Broadly, all
computer computations are performed using addition in different forms.
o A computer assists the research throughout different phases ofresearch
process. It plala a different role at every stage--{onceptual phase, design
and planning phase, empirical phase, analytic phase and dissemination phase.
o The computer is after all a rnan-rnade machine and will only work according
to the software that has been fed in the system.
o Interpretation means drawing inferences from the collected and processed
research data. It establishes continuity in research studies. It acts as a guide
for further research studies and explains the concepts that have been
observed. It is a skill that can be best learned through practice and
experience.
o A researcher must explain the use ofconcepts with the help some theory if
possible. He shouldkeep out extraneous informationbut include relevant
factors while interpreting research studies. The researchers must try and
make sure that the source ofthe collected data is trustworthy and accurate
andthat the datareflects goodhomogeneity. It is important to use proper
statistical methods for analysis. It is important to be aware of and take
precautions against all po s sible sources o f error.
o The researchers also must keep a rvatch over the tendency to over-generalize
or be biased about the results.
o Basically, a research report can be divided into three broad sections: (i)
Preliminary pages-this sectio n carries the t itle, date, acknowledgements,
preface, contents, and a list oftables and illustrations used throughout the
report; (ii) Main text-the main text includes the introduction, statement of
findings and recommendations, results, implications ofthe results and a
sufirmary; and (iii) Closing pages-here the appendices are listed along
with the bibliography of all sources consulted or relevant to the research
sfudy.
o The lay-out is the structure or format ofa research report. It is based on the
universally accepted format provided by the American Psychological
Association manual. Theyhave suggested eight sections under which every
research report has to be written. These are (i) Title, (ii) Abstract, (iii)
Introductioq (iv) Method, (v) Result. (vi) Discussion, (vii) References and
(viii) Appendix (if suitable)

SeU:lnstructional Llaterial 187

Role of Computers in o Research repofis vary in length ancl fomrat. Research reports can be in the
Research, Intetpretation
form ofa technical repofi, popular report, an article, a manuscript or even
and Report Witing
an oral presentation.

. An oral presentation is made for policy makers who wish to have easy and
NOTES
quick access to results of the policy.

5.8 KEY T'ERMS

o Binary system: A systern used by a computer for all its internal
conrnunications.
o CPU: The heart of a computer system which has arithmetical logic systern,
a cache-that stores temporarymemory andcontrol unit.
o Technical report: A detailed report required for record keeping or public
dissemination.
o Popular report: Areport that uses few mathematical or technical terms
and more charts and diagrams, large fonts" subheadings and visuals.

5.9 ANSWERS TO 'CHECK YOUR PROGRESS'

I . A conputer can be used for numerical calculations, assisting in teaching and

learning, financial forecasts, weather tbrecasts, making clinical diagnosis,
providing entertainment like playing games, watching movies and listening
to music, etc.
2. Letus convert the binarynumber 110101 to its decimal equivalent.

Doublingthe leftmostbit we get 2

Adding to it the bit on its right we get 2 + | :3
Doubling againthe nurnber obtainedwe get 6
Adding to it the next bit we get 6 + 0: 6
Againdoublingweget 12
Addingthelastbitwe get12* 1: [3
Doubling the no. obtained we get 26
Adding the next bit we ger 26 + 0 : 26
Doubling the no. obtained we get 52
Adding the next bit we get 52 + I : 53
1 10101 (binary) : 53 (decimal)
3. Let us convertthe decimalnurnber 65 to binarynumber.
Solution: table for conversion of 65 into its binary equivalent'

l8E Self-lnstmctional Material

 Number to be Role of Cotnputers in
Quotient Remainder
Res ear ch, I n t erp r e t a t i on
divided and Report Witing
65 32 I
32 t6 0
16 8 0 NOTES
8 4 0
4 2 0
2 I 0
I 0 I
100000 I (binary) : 65 (decimal)

4. The computer is capable of completing calculations at a very high speed.

5. The central processing unit is responsible for running programs and
applications on the computer.

5.10 QUESTTONS AND EXERCTSES

Short-Answer Questions
1. Define computer hardware and software.
2. Define the binary number systern
3. What is the centralprocessing unit?

4. Name the different types ofreports.

Long-Answer Questions
1. What is a corryuter? Elaborate on its characteristics and various uses.
2. Whyi^s interpretationafi.rndamentalcomponentofresearchprocess?Explain

3. Explain the techniques of interpretation and report writing.

4. Describe the layout ofa research report, covering all inportant points.

5.11 FURTHER READING

Kothari, C.R. 1985. Research Methodology: Methods & Tbchniques. New
Delhi: Wishwa Prakashan.
Singh, A.K. 2008. Tbsts, Measurements and Research Methods in Behovioural
Sciences. New Delhi: Bharati Bhawan.

Self-Instructional Material 189

 NOTES

SeA-Instructional
190 Material
NOTES

Self-Instructional
Material f91
 NOTES

Self-Instruc lional
Material