RESEARCH DESIGN

Memcha Loitongbam
Professor
MIMS, Manipur University
Let’s refresh….
What counts as research?

What qualities does academic research have that differ

from other types of research?
Research?
Question 1:
I want to buy a new laptop.
 I have gathered information about various brands
I have now decided which one to buy.
Comprises you researching various laptops and
assessing them against certain criteria. (part of your
research)
You needed to verify that the laptop you have bought
does indeed meet the set criteria.
You may also want to clarify how you set about
identifying the laptops.
And how this enabled you to make a decision.
Research?
Question 2:
I have gone to the library and researched a certain
topic.
 I have written up everything I have found out
I presented it to my tutor.

Is this research?
Here you are essentially collecting information (can be
part of your research)
Unless you interpret what you have found it remains
lacking.
You always need to know
 why you are collecting information
 what to do with it once you have it
Research?
Question 3:
 I stumbled across a piece of information while looking
for something else.
Although it was a chance finding, I regard this as a lucky
piece of research.

Is this research?

You shouldn't be relying on luck with your research.
 If you found this key piece of information by luck,
you may want to
 question your information search skills
be more systematic and you can guarantee there is
more information out there that you have missed
(since it was a random find).
Academic research follows a set paradigm
This may vary from discipline to discipline
Paradigms guide how we make decisions and carry out
research
Research paradigm
A paradigm is simply a belief system (or theory) that
 guides the way we do things
or more formally establishes a set of practices
can range from thought patterns to action
Research paradigm
According to Guba (1990), paradigms can be
characterised through their:
 ontology (What is reality?)
epistemology (How do you know something?)
methodology (How do you go about finding out?)
These characteristics create a holistic view of how:
 we view knowledge
we see ourselves in relation to this knowledge and the
methodological strategies we use to un/discover it.
 Paradigm, methods and tools
Paradigm Data collection tools
Methods (primarily)
(examples)
Positivist/ Post positivist Quantitative. "Although qualitative Experiments
(e.g. experimental testing) methods can be used within this Quasi-experiments
Theory Testing paradigm, quantitative methods tend to Tests
be predominant . . ." (Mertens, 2005, p. Scales
12)

Interpretivist/ Constructivist Qualitative methods predominate Interviews

. each individual constructs his/her although quantitative methods may also Observations
own reality be utilised. Document reviews
Theory building Visual data analysis

Transformative Qualitative methods with quantitative Diverse range of tools - particular need
e.g. ideas in relation to an ideology and mixed methods. Contextual and to avoid discrimination. Eg: sexism,
historical factors described, especially racism, and homophobia.
as they relate to oppression (Mertens,
2005, p. 9)

Pragmatic Qualitative and/or quantitative methods May include tools from both positivist
not committed to any one system of may be employed. Methods are matched and interpretivist paradigms. Eg
philosophy or reality. to the specific questions and purpose of Interviews, observations and testing and
the research. experiments.
RESEARCH
DESIGN
Research Design
A framework or a blueprint
A foundation
Specifies details

MIMS, M.U.
Research design discusses
Why?
What?
Where?
When?
Way (How)?

MIMS, M.U.
Components of research design
Specify
Design the the
Construct
Construct an
an
Definetype
the
measurement
of Specify
Specify
appropriate formthe
the
appropriate form
information for
for data
research data Develop
sampling
procedure
sampling a plan
process
process
collection
collection
needed and sample size
of data
and sample size
Components
Components
of analysis
of research
research
design
design
Features of a good research design
A good design is often characterized by adjectives
like:
Flexible
Appropriate
Efficient
Economical
 Generally a good research design is one which
Minimizes bias
Maximizes reliability

MIMS, M.U.
Research design classification

Research Design

C onc lu sive

C onc lu sive

Exploratory
Descriptive
Descriptive
Caus al

Caus al

Cross sectional Longitudinal

Cross sectional Longitudinal
 A comparison of Basic Research Designs
CRITERIA EXPLORATORY DESCRIPTIVE CASUAL
Objective: Discover ideas and Describe Determine cause
insights characteristics or and effect
functions relationships
Characteristics:  Flexible  Marked by the  Manipulation of
 Versatile prior formulation of one or more
 Often the front end specific independent
of total research hypotheses variables
design  Preplanned and  Control of other
structured design mediating variables
Methods: Expert surveys Secondary data Experiments
Pilot surveys Surveys
Secondary data Panels
Qualitative Observational
research and other data

MIMS, M.U.
Exploratory Research
is appropriate to any problem about which little is
known
is to develop tentative explanations, not
demonstrating the viability of a given explanation
 is characterized by flexibility in methodology use

MIMS, M.U.
Exploratory Research Design
Gain insights
for developing
an approach

Isolate key variables Establish

and relationships for priorities for
further examination further research

When
Develop Exploratory
hypotheses Research can be
used?

Identify
alternative
courses of
action
Define a
problem more
precisely
Exploratory research method
 Literature search
 Experience surveys
 Analysis of selected cases
Secondary data
Qualitative research

MIMS, M.U.
Example of Exploratory Research Design
Freshbite is a one and half year old e-commerce start-up company
delivering fresh foods as per the order to customer’s doorstep
through its delivery partners.
The company operates in multiple cities.
Since its inception, the company achieved a high sales growth rate.
However, after completion of the first year, the sales started
declining at brisk rate.
Due to lack of historical data, the sales director was confused
about the reasons for this decline in sales.
He prefer to appoint a marketing research consultant to conduct a
study in order to discern the possible reasons rather than making
assumptions
Eg. Contd.
The prime objective of this research was not to figure
out a solution to the declining sales problem
 But rather to identify the possible reasons, such as
poor quality of products and services
 competition
 or ineffective marketing,
and to better understand the factors affecting sales
Once these potential causes are identified, the
strength of each reason can be tested using causal
research
Conclusive Research Design-
Descriptive Research
Descriptive Research
Purpose
Describe the characteristics of relevant
groups/samples/population
Estimate proportions in specified population
Make specific predictions

MIMS, M.U.
Continued…
Characterized by prior formulation of hypothesis
often follows exploratory research
Information needed is clearly defined, preplanned
and structured
Often uses rigid methods
Typically based on large representative samples
Frequently use pilot studies
Requires clear specification of the Ws of research
 who, what, when, where, why and way

MIMS, M.U.
Descriptive research question examples
How has the Stock Market changed over the past 20 years?
Do customers of company X prefer product X or product Y?
What are the main genetic, behavioural and morphological
differences between wildcats and domestic cats?
What are the most popular online news sources among
under-18s?
How prevalent is disease A in population B?
What are the most significant risks in nuclear energy
production?
Methods of Descriptive Research

Secondary data analyzed in a quantitative as

opposed to a qualitative manner
Surveys
Panels
Observational and other data
 Classifications of descriptive studies

Cross-sectional Sample Survey

Descriptive studies
True Panel
Longitudinal

Omnibus
Panel

MIMS, M.U.
Cross sectional descriptive studies
The study takes place at a single point in time

It does not involve manipulating variables

It allows researchers to look at numerous characteristics at once (age,

income,gender,etc.)

Often used to look at the prevailing characteristics of a given population

Example: Cross-sectional
A researcher might collect cross-sectional data on
past smoking habits and current diagnoses of lung
cancer
This type of study cannot demonstrate cause-and-
effect,
It can provide a quick look at correlation that may
exist at a particular point.
Longitudinal descriptive
Include panel i.e. a final sample of elements, which
are repeatedly measured over time
Panel members are relatively constant over time
Main disadvantage of panels is that they are non
representative

MIMS, M.U.
Longitudinal Descriptive Research

• True panel: repeatedly measured on the same

variable
• Omnibus panel: repeatedly measured, but on
differing variables

MIMS, M.U.
 Cross Sectional vs. Longitudinal Designs
Figure 3.6 Cross Sectional vs.
Longitudinal Designs

Cross Sectional
Design Sample
Surveyed at T1

Same Sample
Longitudinal Design Sample also Surveyed
Surveyed at T1 at T2

Time T1 T2
Major Concerns in Designing, Conducting and
Interpreting Descriptive-Survey Research
A. Are the Data (Findings) Valid (True)?
 Source of Invalidity of Data = MEASUREMENT ERROR

B.To What Population(s) can the findings be generalized?

External Validity of Descriptive-Survey Research
 Sources of External Invalidity = SAMPLING ERROR
 Non-representative sample
 non-probability sample

MIMS, M.U.
 Causal Research

MIMS, M.U.
Casual research
A change in one variable brings about a change in
another variable
Purpose
To understand which variables are the cause (independent
variables) and which variables are the effect (dependent
variables) of a phenomenon
To determine the nature of the relationship between the
casual variables and the effect to be predicted
Requires a planned and structured design

MIMS, M.U.
Causal Research

Causality tested through

• Lab experiments
- All variables are controlled
- Influence of external variables minimized
-Internal validity maximized
• Field tests
- Most credible results
- Influence of ‘real world’ settings taken into account
-External validity maximized

MIMS, M.U.
INTERNAL VALIDITY: indicates whether the
independent variable was the sole cause of the change in
the dependent variable
 EXTERNAL VALIDITY: involves the extent to which the
results of a study can be generalized (applied) beyond the
sample
VALIDITY The internal validity of any research
undertaking is strengthened by the correct choice of
research design and the soundness and appropriateness
of decisions that pertain to sampling, instrumentation,
data collection and analysis.
THREATS TO EXPERIMENTAL VALIDITY
History
Maturation
Testing
Instrumentation
Selection
Regression
 Experimental mortality
Classification of experimental designs
Experimental Designs

Pre- True Quasi- Statistical

experimental Experimental Experimenta
Pretest –Posttest
l
Time Series
One-Shot Case Randomized
Control Group
Study Multiple Blocks
Posttest – Only Time Series
One Group Latin Square
Control Group
Pretest-Posttest Factorial
Solomon - Four
Static Group Group

MIMS, M.U.
Pre-experimental case study
One-shot case study (after only design)
Do not employ randomization procedure to control
extraneous factors
A single group of test unit is exposed to a treatment
(X)
Single measurement on the dependent variable is
taken (O1)
No control on extraneous variables
Symbolically: X O1

MIMS, M.U.
ONE-SHOT CASE STUDY DESIGN
Example : Effects of extra coaching on non performing
students
Experi Treatm Post
mental test(O
ent(X)
Group
Extra 1)
Non
perfor coachi Obser
mers ng vation
One-Group Pretest –Posttest Design
Symbolized as
O1 x O2
A group of test unit is measured twice
No control group
First a pretreatment measure is taken (O1)
Group is then exposed to treatment (X)
Post treatment measure is taken (O2)
Treatment effect is computed as O2 – O1
Extraneous variables are largely uncontrolled

MIMS, M.U.
Example: You want to determine whether praising
primary school children makes them do better in
Mathematics.
Experi
mental Treat Post
Group ment test
Pretest
(O1)
(X) (O2)
Static Group Design
Two-group experimental design
Experimental group (EG) is exposed to treatment (X)
Control group (CG) is not exposed to treatment
Measurements on both group are made only after the
treatment
Test units are not assigned at random
Symbolically
EG: X O1
CG: O2
Treatment effect -> O1 – O2

MIMS, M.U.
Example: You want to determine whether a
motivation workshop makes students perform
better.
Group A (EG) Treatment (X) Post test (O1)

Group B (CG) No treatment Post test (O2)

Treatment effect -> O1 – O2

True Experimental Design
Distinguishing feature is randomization
- test units randomly assigned to experimental
groups
- treatments randomly assigned to experimental
groups

MIMS, M.U.
True Experimental Design
Posttest – Only Control Group Design
No pretest measure is taken
EG exposed to the treatment
CG not exposed
Random assignment made
Symbolically:
EG : R X O1
CG : R O2
Treatment effect (TE) = O1 – O2

MIMS, M.U.
Posttest – Only Control Group Design
Pretest –Posttest control Group design
EG exposed to treatment
CG not exposed
Test units randomly assigned
Pretest measurement taken
Symbolically:
EG: R O1 X O2
CG: R O3 O4
Treatment effect (TE) ->
(O2 – O1) – (O4 – O3)

MIMS, M.U.
Pretest –Posttest control Group design
Solomon Four-Group design
Overcomes the limitations of the pretest – posttest
control group and posttest only control group
design
The population is randomly divided into four
samples
Two of the groups are experimental group
Two groups experience no experimental
manipulation
Two groups receive pretest and a post test
Two groups receive only a post test

MIMS, M.U.
Continued…
R O1 T O2

R O1 O2
Population R Sample
R T O2

R O2

MIMS, M.U.
Quasi Experimental Design
Researcher can control when measurement are
taken and on whom they are taken
Researcher lacks control over the scheduling of
the treatments
Unable to expose test units to the treatment
randomly
Can be used in cases when true experimentation
cannot be conducted
Quick and less expensive

MIMS, M.U.
Ex post facto research design
Also known as "after the fact" research,
 an ex post facto design is considered quasi-
experimental because the subjects are not randomly
assigned –
 they are grouped based on a particular characteristic
or trait.
For example, a researcher is interested in how weight
influences self-esteem levels in adults.
So the participants would be separated into differing
groups (underweight, normal weight, overweight) and
their self esteem levels measured.
This is an ex post facto design because a pre-existing
characteristic (weight) was used to form the groups.
Quasi Experimental Design: Forms
Time series design
 Includes periodic measurements on depended variable
 Treatment is administered by the researcher or occurs
naturally
 After the treatment periodic measurements are
continued

MIMS, M.U.
Continued…
Multiple Time series Design
 A time series design that includes another group of test
units to serve as a control group
 Improvement over the simple time series

MIMS, M.U.
Statistical Design
The effects of more than one independent variable can
be measured
Specific extraneous variables can be statistically
controlled
Economical designs can be formulated

MIMS, M.U.
Completely Randomized Design
Used to investigate the effect of one independent variable
on the dependent variable
Eg. Sales influenced by price - at which price level sale is
highest
Dependent Independent

Independent variable (IV) to be measured in nominal scale

(should have a number of categories – eg. Low, medium,
high)
Each category of IV considered as treatment –i.e. Low,
medium and high
All test units are treated alike – eg. Stores
Completely Randomized Design
Test units (stores here) are randomly assigned to the
three treatment levels ( 3 price levels)
 the average sales for each price level is computed and
examined to find any significant differences
 Statistical technique - ANOVA
One limitation is –effect of extraneous variable ( size
of store, competitor’s price, price of substitute) not
taken in account
Randomized Block Design
Overcomes the limitation of Completely Randomized
Design – separates the influence of extraneous variable
on dependent variable
Subjects are divided into groups known as block
Subjects are relatively homogeneous
Useful only when there is only one major external
variable, e.g.: sales, store size, income, etc

MIMS, M.U.
Eg.
Independent Variable- Price level: low , medium and
high
Test Units - Stores now considered different: small,
medium and large
Extraneous variable (size of stores) is treated as
different blocks
Treatments are randomly assigned to the blocks in
such a way that each treatment appears in each block
at least once
Statistical technique- ANOVA
Latin Square Design
Used to control two non-interacting external variables
as well as the independent variable
External and internal variables divided into equal
number of blocks or levels
Eg. Suppose the interest is to study the influence of
price(treatment) on sales
 let there be 3 l3vels of price category- low(X 1),
medium(X2), high(X3)
Sales could be affected by 2 extraneous variables- store
size and type of packaging
MIMS, M.U.
Store size – (1), (2), (3)
Type of packaging- I, II, III
Treatments are assigned randomly
 Helps to measure statistically effect of treatment on
dependent variable and measurement of error
resulting from two extraneous variables
Store size Packaging
I II III
1(Small) X1 X2 X3
2(Medium) X2 X3 X1
3(Large) X3 X1 X2
Example: Driver frustration under low , medium and high density traffic condition
and traffic flow controlled by a police officer or traffic signal was investigated. The
measure of frustration was the number of horns honked by the drivers before
getting the green light at a controlled intersection. The mean number of horns
honked in each condition were:
Research Design- Potential Sources of Error

 A good research design attempts to control the various

sources of error
 Total error is the variation between the the true mean value
in the population and the observed mean value obtained in
the marketing research study.
 There are two types of errors:
 Random sampling: an imperfect representation of the
population.
 Non sampling error: attributed to sources other than
sampling.

MIMS, M.U.
 Potential Sources of Error in Research Designs
Total Error

Random Sampling Non-sampling

Error Error

Response Non-response
Error Error

Researcher Error Interviewer Error Respondent

Error

Surrogate Information Error Respondent Selection Error Inability Error

Measurement Error Questioning Error Unwillingness Error
Population Definition Error Recording Error
Sampling Frame Error Cheating Error
Data Analysis Error
“Research is to see what everybody else has seen,
and to think what nobody else has thought.”
Albert Szent- Gyorgyi

THANK YOU