4585_743_24_973_45_RESEARCH DESIGN (1)

8/6/2019 4585_743_24_973_45_RESEARCH DESIGN (1)

http://slidepdf.com/reader/full/45857432497345research-design-1 1/40

RESEARCH DESIGN

� Framework/Blueprint for research plan of action, giving a general statement of themethods to be used.

� Specifies the details.� Components (Specification of): ± Type of Design

± Information needed

± Measurement & Scaling Techniques

± Questionnaire construction & pretesting

± Sampling process & size

± Data Analysis Plan

8/6/2019 4585_743_24_973_45_RESEARCH DESIGN (1)


CLASSIFICATION

� EXPLORATORY RESEARCH

± To provide insight into, & an

understanding of, the problem (s) faced.

� CONCLUSIVE RESEARCH

± To assist in determining, evaluating, &

selecting the best course of action to take

in a given situation.

± DESCRIPTIVE RESEARCH

± CAUSAL RESERACH

8/6/2019 4585_743_24_973_45_RESEARCH DESIGN (1)


EXPLORATORY RESEARCH

DESIGN� To explore or search through the problem/situation for sharper focus of situation

� PURPOSES:

± Identification of problem ± Precise formulation of the problem

± Identify alternative courses of action

± Develop hypothesis

± Isolate key variables & their relationships� Not a conclusive study so design is flexible

& informal

8/6/2019 4585_743_24_973_45_RESEARCH DESIGN (1)


EXPLORATORY RESEARCH

DESIGN: METHODS� SURVEY OF SECONDARY DATA

� EXPERIENCE SURVEY

� QUALITATIVE RESEARCH� PILOT SURVEY

8/6/2019 4585_743_24_973_45_RESEARCH DESIGN (1)


EXPLORATORY RESEARCH DESIGN:

SURVEY OF SECONDARY DATA

� Existing Data or Literature

� Located & collected rapidly, easily & inexpensively

� A prerequisite to the collection of primary data

� May not provide solution but provide direction

� CLASSIFICATION:

± Internal

� Ready to Use

� Requires Further Processing

± External

� Published Material

� Computerized Databases

� Syndicated Services

8/6/2019 4585_743_24_973_45_RESEARCH DESIGN (1)


EXPERIENCE SURVEY

� Obtaining information from Experienced &

Knowledgeable persons

� Judgement Sampling + Referral Sampling

� Unstructured & Informal

8/6/2019 4585_743_24_973_45_RESEARCH DESIGN (1)


QUALITATIVE RESERACH

� An unstructured exploratory researchmethodology based on small sample thatprovides insights & understandings

� Quantitative research must be preceded byappropriate qualitative research

� Not possible/desirable to use structured/formalmethod: ± People are unwilling/unable to answer

± Unable to provide accurate answer toquestions that tap their subconscious

± Values, emotional drives, motivations aredisguised

� Not conclusive & should not be generalized.

8/6/2019 4585_743_24_973_45_RESEARCH DESIGN (1)


8/6/2019 4585_743_24_973_45_RESEARCH DESIGN (1)


FOCUS GROUP INTERVIEWS

� Joint informal interview in the form of freeflowing group-discussions

� To gain insight by listening to people from

target group� Most important qualitative research

technique

� Group size- 8 to 12, Group composition-

homogeneous, prescreened, Time duration-1 to 3 hrs

� Role of Moderator

8/6/2019 4585_743_24_973_45_RESEARCH DESIGN (1)


DEPTH INTERVIEWS

� Unstructured direct, personal interview, on

one to one basis

� Single respondent is probed to uncover

underlying motivations, beliefs, attitudes &feelings on a problem/situation.

� Interviewing Techniques:

± Laddering: Issue characteristics User characteristics ± Hidden Issue Questioning

± Symbolic Analysis: Comparing with opposites.

8/6/2019 4585_743_24_973_45_RESEARCH DESIGN (1)


PROJECTIVE TECHNIQUES

� Unstructured, indirect form of questioning,encouraging respondents to project their underlying motivations, beliefs, attitudes etc

� Respondents are asked to interpret the behavior of others rather than own

� Classification:

± Association Techniques: Word Association

± Completion Techniques: Sentence Completion,Story Completion

± Construction Techniques: Picture Response,Cartoon Tests

± Expressive Techniques: Role Playing, Third-

Person Technique

8/6/2019 4585_743_24_973_45_RESEARCH DESIGN (1)


PILOT STUDY

� Small scale exploratory research survey

� Uses Sampling but rigid standards are not

applied.

� Primary data, Qualitative Analysis.

� To identify and eliminate potential problems

in research administration

8/6/2019 4585_743_24_973_45_RESEARCH DESIGN (1)


DESCRIPTIVE RESEARCH

DESIGN� To know the characteristics of certain groups

or situations

� To facilitate description & inference building

about parameters & their relationships.� Uncover causal relationship but not

established

� Assumes prior knowledge about the problem

situation� Structured & formal design

� Methods: Quantitative Analysis of Secondarydata, Surveys, Panels, Observational Study

8/6/2019 4585_743_24_973_45_RESEARCH DESIGN (1)


DESCRIPTIVE RESEARCH

DESIGN: CLASSIFICATION� LONGITUDINAL RESEARCH / PANEL

RESEARCH DESIGN

� CROSS-SECTIONAL RESEARCHDESIGN

± SINGLE CROSS-SECTIONAL DESIGN

± MULTIPLE CROSS-SECTIONALDESIGN

� COHORT ANALYSIS

8/6/2019 4585_743_24_973_45_RESEARCH DESIGN (1)


LONGITUDINAL RESEARCH

� Based on PANEL DATA & PANEL METHOD

� Continual or periodic collection of information from fixed sample of respondents

� Related to repeated measurement of thesame variable (s)

� To detect changes because of repeatedmeasurements

� OMNIBUS PANEL: ± Information collected from fixed sample

may vary over time or may vary betweenindividuals

8/6/2019 4585_743_24_973_45_RESEARCH DESIGN (1)


LONGITUDINAL RESEARCH(Cross-Sectional Data)

Brand

Purchased

Period 1 Survey Period 2 Survey

Brand A 200 200

Brand B 300 300

Brand C 500 500

Total 1000 1000

8/6/2019 4585_743_24_973_45_RESEARCH DESIGN (1)


LONGITUDINAL RESEARCH(Longitudinal Data: Brand-Switching Matrix)

Brand Purchased

in Period 1

Brand Purchased in Period 2

Brand A Brand B Brand C Total

Brand A 100 50 50 200

Brand B 25 100 175 300

Brand C 75 150 275 500

Total 200 300 500 1000

8/6/2019 4585_743_24_973_45_RESEARCH DESIGN (1)


CROSS-SECTIONAL DESIGN

� Aimed at taking one time stock of thesituation

� Collection of information only once

� Most frequently used DescriptiveDesign.

� SINGLE CROSS-SECTIONAL DESIGN:

± One sample of respondents &information is obtained from the sampleonly once

8/6/2019 4585_743_24_973_45_RESEARCH DESIGN (1)


CROSS-SECTIONAL DESIGN

� MULTIPLE CROSS-SECTIONAL DESIGN:

± Two or more samples of respondents &information is obtained from each sample

only once (Information at different times) ± Allows comparison at the aggregate level

but not at the individual level

± COHORT ANALYSIS:

� Series of surveys at appropriate timeinterval from the groups of respondents(Cohorts) who experience the sameevent within the same time interval

8/6/2019 4585_743_24_973_45_RESEARCH DESIGN (1)


COHORT ANALYSIS:EXAMPLE(Consumption of Soft Drink by various Age Groups)

Age 1980 1990 2000 2010

10-19 53 63 73 81

20-29 45 61 76 76

30-39 34 47 68 71

40-49 23 41 59 68

50+ 18 29 50 52

8/6/2019 4585_743_24_973_45_RESEARCH DESIGN (1)


CAUSALITY

� The basis of classification in betweenExperimental & Non-Experimental designs

� Conditions for making Causal Statement:

± Concomitant Variation ± Time Sequence

± Absence of Extraneous Factor (s)

� Control of Extraneous Variables:

± Physical Control

± Randomization

± Experimental Design

8/6/2019 4585_743_24_973_45_RESEARCH DESIGN (1)


EXPERIMENTATION

� Manipulation of variable (s) and control

of other variables and examining their

effect

ELEMENTS

INPUTS OUTPUT

EXPLAN ATORY & DEPE NDENTEXTRANEOUS VARIABLE

VARIABLES

TEST UNITS

8/6/2019 4585_743_24_973_45_RESEARCH DESIGN (1)


VALIDITY OF EXPERIMENT

� INTERNAL VALIDITY

� EXTERNAL VALIDITY

� Factors affecting Validity: ± HISTORY

± MATURATION

± TESTING EFFECT

± INTRUMENTATION

± SELECTION BIAS

± TEST UNIT MORTALITY

8/6/2019 4585_743_24_973_45_RESEARCH DESIGN (1)


EXPERIMENTAL DESIGN

� Stronger & more reliable basis for the

existence of Causal Relationship

� All extraneous variables are eliminated

through the Use of Randomization & ControlGroup (s)

8/6/2019 4585_743_24_973_45_RESEARCH DESIGN (1)


TYPES OF EXPERIMENTAL DESIGN

AFTER ONLY WITH ONE CONTROL GROUP

BEFORE-AFTER WITH ONE CONTROL GROUP

EX-POST FACTO DESIGN

SIMULATED BEFORE AFTER DESIGN

SOLOMON¶S FOUR GROUPS DESIGN

MULTIPLE TIME SERIES DESIGN

C.R.D.

R.B.D.

L.S.D.

FACTORIAL DESIGN

8/6/2019 4585_743_24_973_45_RESEARCH DESIGN (1)


AFTER-ONLY WITH ONE

CONTROL GROUP

� Treatment Effect = O1 - O2

� Very widely used in marketing

� Control over µTesting¶ & µInstrument Effect¶

� No pre- observations so large samples are

required

EXPERIMENTAL GROUP: R X O1

CONTROL GROUP : R O2

8/6/2019 4585_743_24_973_45_RESEARCH DESIGN (1)


BEFORE-AFTER WITH ONE

CONTROL GROUP

� TREATMENT EFFECT = (O2 O1) (O4 O3)

� Treatment Effect contains µTesting Effect¶

� Effective relative comparison

EXPERIMENTAL GROUP: R O1 X O2

CONTROL GROUP : R O3 O4

8/6/2019 4585_743_24_973_45_RESEARCH DESIGN (1)


EX-POST FACTO DESIGN

� TREATMENT EFFECT = O1 O2

� Used in Advt. Study & study of Commercial

Media

� No µTesting Effect¶ & more realistic study

EXPERIMENTAL GROUP: X R O1


8/6/2019 4585_743_24_973_45_RESEARCH DESIGN (1)


SIMULATED BEFORE-AFTER

DESIGN


� No µ Testing Effect ¶

� Used in Advt. Research

EXPERIMENTAL GROUP: R X O2


8/6/2019 4585_743_24_973_45_RESEARCH DESIGN (1)


SOLOMON¶S FOUR ± GROUP

DESIGN

� Also known as FOUR GROUP SIX STUDY

DESIGN

EXPERIMENTAL GROUP 1 : R O1 X O2

CONTROL GROUP 1 : R O3 O4

EXPERIMENTAL GROUP 2 : R X O5

CONTROL GROUP 2 : R O6

8/6/2019 4585_743_24_973_45_RESEARCH DESIGN (1)


SOLOMON¶S FOUR ± GROUP

DESIGN� Combination of µBEFORE-AFTER WITH ONE

CONTROL GROUP¶ & µAFTER-ONLY WITH ONECONTROL GROUP¶ Designs. So also known as µTHEIDEAL CONTROL EXPERIMENT¶

� Not very commonly used because of high cost &high efforts


� TESTING EFFECT = (O2 O1) (O4 O1 + O3)

2� EXTRANEOUS EFFECT = O6 ( O1 + O3 )

2

8/6/2019 4585_743_24_973_45_RESEARCH DESIGN (1)


MULTIPLE TIME SERIES DESIGN

EXPERIMENTAL GR: O1 O2 O3«..X«. O4 O5 O6

CONTROL GR : O1 O2 O3«. ...«. O4 O5 O6

� Extension of BEFORE-AFTER WITH ONE

CONTROL GROUP DESIGN

� Also known as µCONTINUOUS DIARY PANEL

DESIGN� Better analysis of trend

� Higher cost.

8/6/2019 4585_743_24_973_45_RESEARCH DESIGN (1)


MANIPULATION / CONTROL OF MORE

THAN ONE LEVELS OF TREATMENTS

COMPLETELY RANDOMIZEDDESIGN (C.R.D.)

� To study the effect of different levels/categories (Treatment) of explanatory

variable on dependent variable� Treatment are assigned completely at

random to experimental units

8/6/2019 4585_743_24_973_45_RESEARCH DESIGN (1)


COMPLETELY RANDOMIZED

DESIGN (C.R.D.)

� Uses only the principles of Replication

& Randomization but ignores Local

Control

� Useful only in small preliminary

experiments where experimental units

are homogeneous

� Analysis - ONE WAY CLASSIFICATION

8/6/2019 4585_743_24_973_45_RESEARCH DESIGN (1)


RANDOMIZED BLOCK DESIGN

(R.B.D.)

� It is possible to separate out the effectof one extraneous factor

� All three principles: RANDOMIZATION,

REPLICATION & LOCAL CONTROL areused.

� There are µm¶ treatments, each being

replicated µn¶ times. So, N= m.nexperimental units & units are nothomogeneous.

8/6/2019 4585_743_24_973_45_RESEARCH DESIGN (1)



(R.B.D.)

� R.B.D. consists of two steps :

STEP 1

� The experimental units are allocated toµn¶ homogeneous BLOCKS, number of units within a block = number of treatments.

STEP 2� Assign treatments at random to units of

a block.

8/6/2019 4585_743_24_973_45_RESEARCH DESIGN (1)



(R.B.D.)

� By restricting randomization to a blockcontrol over extraneous factor (error) isobtained.

� Most popular but block size ( no. of treatments ) should be small o.w.blocks may not be internallyhomogeneous.

� Analysis: TWO WAY CLASSIFICATIONof the data & may be used to examinedifferent levels of two factors.

8/6/2019 4585_743_24_973_45_RESEARCH DESIGN (1)


LATIN SQUARE DESIGN (LSD)

� Principle of LOCAL CONTROL is used

by grouping the units in two ways:

Rows & Columns

� Each way corresponding to a source of

variation among the units.

� It allows the control & measurement of

effect of two extraneous variables on

dependent variable.

8/6/2019 4585_743_24_973_45_RESEARCH DESIGN (1)



� Condition : Two extraneous variables

have same number of categories as that

of treatments. So number of treatments

= number of replications = m

� Total number of experimental units = m2

( arranged in m rows & m columns,

according to two sources of variation )

8/6/2019 4585_743_24_973_45_RESEARCH DESIGN (1)



� The m treatments are allotted to these m2

units at random s.t. each treatment occursonce & only once in each row & eachcolumn. This is incomplete 3 ± way layout

C1 C2 C3 C4 C5

Ex. R1 A E D C B

R2 D B A E C

R3 B A C D E

R4 C D E B A

R5 E C B A D

4585_743_24_973_45_RESEARCH DESIGN (1)

Documents