Quantitative Research Design

Md. Zakir Hossain, Ph.D.

Professor
Urban and Rural Planning Discipline
Khulna University, Khulna 9208
Bangladesh
What is a research design?

A research design provides a framework for the

collection and analysis of data. A choice of research
design reflects decisions about the priority being
given to a range of dimensions of the research
process. These include the importance attached to:
 expressing causal connections between variables;
 generalizing to larger groups of individuals than those
actually forming part of the investigation;
 understanding behaviour and the meaning of that
behaviour in its specific social context;
 having a temporal (that is, over time) appreciation of
social phenomena and their interconnections.
What is a research method?

A research method is simply a technique for

collecting data. It can involve a specific
instrument, such as a
self-completion questionnaire or a structured
interview schedule, or participant observation
whereby the
researcher listens to and watches others.
Experimental Design
• Experiments are used to study causal relationships. You manipulate one or more
independent variables and measure their effect on one or more dependent
variables.
• Experimental design create a set of procedures to systematically test a hypothesis. A
good experimental design requires a strong understanding of the system you are
studying.
• seeks to determine if a specific treatment influences an outcome. The researcher
assesses this by providing a specific treatment to one group and withholding it from
another and then determining how both groups scored on an outcome.
• The researcher has control over the experiment in terms of:
 Who is being researched (subjects randomly assigned).
 What is being researched.
 When the research is to be conducted.
 Where the research is to be conducted.
 How the research is to be conducted.
Experimental Design

Designing a sound experiment involves five essential steps:

• Identify your variables and examine their relationship – Clearly define your
independent (manipulated) and dependent (measured) variables.
• Formulate a clear, testable hypothesis – Your hypothesis should make a
specific prediction about the effect of one variable on another.
• Develop experimental treatments – Create variations of your independent
variable to observe how they affect outcomes.
• Determine how to assign participants to groups – Choose between a
between-subjects design (different participants in each group) or a within-
subjects design (the same participants experience all conditions).
• Define how you’ll measure your dependent variable – Ensure your
measurement method is valid, reliable, and suited to the hypothesis.
 Classical Experimental design
8 months
T1
Exp T2 Experimental
Obs1 IQ
Teacher Obs2 IQ Group
expectancies Spurters
Random
assignment

No Exp Control
Obs3 IQ Teacher Obs4 IQ Group
expectancies Non-Spurters

Obs An observation made in relation to the dependent variable; there may well be two or more observations, IQ tests
and reading grades before and after the experimental manipulation

Exp The experimental treatment (an independent variable) such as teachers’ expectation. No exp. refers to the
absence of an experimental treatment and represents the experience of control group

T = The timing of the observations made in relation to the dependent variable

Experimental Design
Step 1: Define your variables

You should begin with a specific research question. We will work with two
research question examples, one from health sciences and one from ecology:

Example question 1: Phone use and sleep

You want to know how phone use before bedtime affects sleep patterns.
Specifically, you ask how the number of minutes a person uses their phone
before sleep affects the number of hours they sleep.

Example question 2: Temperature and soil respiration

You want to know how temperature affects soil respiration. Specifically, you
ask how increased air temperature near the soil surface affects the amount of
carbon dioxide (CO2) respired from the soil.
Experimental Design
To translate your research question into an experimental hypothesis, you need to define the main
variables and make predictions about how they are related.
Research question Independent variable Dependent variable

Phone use and sleep Minutes of phone use before sleep Hours of sleep per night

Temperature and soil respiration Air temperature just above the soil CO2 respired from soil
surface

Then you need to think about possible extraneous and confounding variables and consider how you
might control them in your experiment.
Research question Extraneous variable How to control

Phone use and sleep Natural variation in sleep Control statistically: measure the average difference
patterns among individuals. between sleep with phone use and sleep without
phone use rather than the average amount of sleep
per treatment group.

Temperature and soil Soil moisture also affects Control experimentally: monitor soil moisture and add
respiration respiration, and moisture can water to make sure that soil moisture is consistent
decrease with increasing across all treatment plots.
temperature.
Experimental Design

Here we predict that the amount of phone use Here we predict that increasing temperature
will have a negative effect on hours of sleep and will increase soil respiration and decrease
predict an unknown influence of natural variation soil moisture, while decreasing soil moisture
on hours of sleep. will lead to decreased soil respiration.
Step 2: Write your hypothesis
Now that you have a strong conceptual understanding of the system you are studying, you should
be able to write a specific, testable hypothesis that addresses your research question.

Research question Null hypothesis (H0) Alternate hypothesis (H1)

Phone use and sleep Phone use before sleep does not Increasing phone use before sleep
correlate with the amount of sleep a leads to a decrease in sleep.
person gets.
Temperature and soil respiration Air temperature does not correlate Increased air temperature leads to
with soil respiration. increased soil respiration.

The next steps will describe how to design a controlled experiment. In a controlled
experiment, you must be able to:

• Systematically and precisely manipulate the independent variable(s).

• Precisely measure the dependent variable(s).
• Control any potential confounding variables.
Step 3: Design your experimental treatments
How you manipulate the independent variable can affect the experiment’s external
validity – that is, the extent to which the results can be generalized and applied to
the broader world.
Soil-warming experiment
You can choose to increase air temperature:
just slightly above the natural range for your study region. over a wider range of
temperatures to mimic future warming. over an extreme range that is beyond any
possible natural variation.

Phone-use experiment
You can choose to treat phone use as:
• a categorical variable: either as binary (yes/no) or as levels of a factor (no phone
use, low phone use, high phone use).
• a continuous variable (minutes of phone use measured every night).
Step 4: Assign your subjects to treatment groups
First, you need to consider the study size: how many individuals will be included in
the experiment? In general, the more subjects you include, the greater your
experiment’s statistical power, which determines how much confidence you can
have in your results.

Then you need to randomly assign your subjects to treatment groups. Each group
receives a different level of the treatment (e.g. no phone use, low phone use, high
phone use).

You should also include a control group, which receives no treatment. The control
group tells us what would have happened to your test subjects without any
experimental intervention.

When assigning your subjects to groups, there are two main choices you need to
make:
• A completely randomized design vs a randomized block design.
• A between-subjects design vs a within-subjects design.
Randomization
An experiment can be completely randomized or randomized within blocks (aka strata):
• In a completely randomized design, every subject is assigned to a treatment group at
random.
• In a randomized block design (aka stratified random design), subjects are first grouped
according to a characteristic they share, and then randomly assigned to treatments within
those groups.
Research question Completely randomized Randomized block design
design

Phone use and sleep Subjects are all randomly assigned a Subjects are first grouped by age,
level of phone use using a random and then phone use treatments are
number generator. randomly assigned within these
groups.

Temperature and soil respiration Warming treatments are assigned to Soils are first grouped by average
soil plots at random by using a rainfall, and then treatment plots are
number generator to generate map randomly assigned within these
coordinates within the study area. groups.
Between-subjects vs. within-subjects: Example
Between-subjects design (Independent measures design):In this design, each participant
is assigned to only one group or condition.
• Example: If you're testing two different types of diets (Diet A and Diet B) to see which
helps people lose more weight, you divide participants into two groups. One group
follows Diet A, and the other follows Diet B. Each person experiences only one diet.
Good for: Avoiding learning or carryover effects.
Downside: Needs more participants, and individual differences between groups may
affect results.

Within-subjects design (Repeated measures design):In this design, the same participants
take part in all conditions of the experiment.
• Example: If you're testing the effect of two different diets, every participant tries both
Diet A and Diet B, one after the other (with a break in between). Their weight loss is
compared across the two diets.
Good for: Reducing variability since each person is compared to themselves.
Downside: Can cause order or fatigue effects (e.g., they may lose more weight just
because they tried harder the first time).
Between-subjects vs. within-subjects: Example

Research Between-subjects (independent Within-subjects (repeated

question measures) design measures) design

Phone use and sleep Subjects are randomly assigned a level of phone Subjects are assigned consecutively to
use (none, low, or high) and follow that level of zero, low, and high levels of phone use
phone use throughout the experiment. throughout the experiment, and the order
in which they follow these treatments is
randomized.

Temperature and soil Warming treatments are assigned to soil plots at Every plot receives each warming
respiration random and the soils are kept at this temperature treatment (1, 3, 5, 8, and 10C above
throughout the experiment. ambient temperatures) consecutively over
the course of the experiment, and the
order in which they receive these
treatments is randomized.
Step 5: Measure your dependent variable
• Finally, you need to decide how you’ll collect data on your dependent variable
outcomes. You should aim for reliable and valid measurements that minimize
research bias or error.

• Some variables, like temperature, can be objectively measured with scientific

instruments. Others may need to be operationalized to turn them into
measurable observations.

Phone use experiment

In your experiment about phone use and sleep, you could measure your
dependent variable in one of two ways:
• Ask participants to record what time they go to sleep and get up each day.
• Ask participants to wear a sleep tracker.

How precisely you measure your dependent variable also affects the kinds of
statistical analysis you can use on your data.
Principles of RCTs in Social Science Research

Clear Research Question & Hypothesis

• Define a precise and testable hypothesis (e.g., Does vocational training
improve employment outcomes for women in urban slums?).
• Specify outcome variables (income, health, education, behaviour change).

Randomisation
Participants (individuals, households, schools, communities, or even villages)
are randomly assigned to treatment and control groups.
Ensures comparability and controls for both known and unknown confounding
factors.
Randomisation can be:
• Simple randomisation (lottery, random number generator)
• Cluster randomisation (whole communities or schools)
• Stratified randomisation (balancing groups based on key
characteristics, e.g., gender, age, income).
Principles of RCTs in Social Science Research

Control Group / Counterfactual

A valid comparison group is essential to measure what would have happened
without the intervention.
Control can be:
• Pure control (no intervention)
• Placebo/alternative intervention (receives some neutral activity)
• Wait-list control (receives the intervention later).

Blinding (where possible)

Harder in social science than in drug trials.
Full blinding is rare, but bias can be reduced by:
• Blinding data collectors and outcome assessors.
• Ensuring participants don’t know the full details of expected outcomes.
• Using objective outcome measures (e.g., school attendance records,
employment data, voting turnout).
Principles of RCTs in Social Science Research

Standardised Intervention Delivery

• The intervention (training, cash transfer, awareness campaign, etc.) should
be delivered consistently across all treatment units.
• Manuals, protocols, and training for facilitators ensure fidelity.

Outcome Measurement
• Outcomes must be relevant, reliable, and measurable (e.g., income,
health behaviour, educational performance, empowerment indices).
• Use both quantitative (surveys, records, behavioural data) and
qualitative (interviews, FGDs) methods to capture impact.
• Where possible, prefer objective measures over self-reports to minimise
bias.
Principles of RCTs in Social Science Research
Intention-to-Treat (ITT) Principle
• Analyse participants in the groups they were originally assigned, regardless of whether
they fully complied.
• Maintains the benefits of randomisation and avoids attrition bias.

Ethics & Informed Consent

• Participants must give informed consent, understanding risks and benefits.
• Special care for vulnerable populations (poor, migrants, children, women, PwDs).
• Control groups must not be deprived of basic rights or essential services.

Quality Assurance & Monitoring

• Continuous supervision to ensure the intervention is delivered as designed.
• Regular audits and pre-tests (pilots) to refine design and tools.
Transparent Reporting & Replication
• Follow reporting standards (e.g., CONSORT for social experiments).
• Clearly state methodology, randomisation process, attrition rates, and limitations.
• Encourage replication in different contexts to test external validity.
Citizen Perception Survey on Service Delivery
and Institutional Efficiency of Upazila Parishad
and Union Parishad

The overall objective of the baseline study is to

engage citizens, service users, union parishad and
upazila parishad, and service providers in a
meaningful assessment and debate on the quality of
service and governance at the local level.

Specific objectives of the survey are:

Obtaining citizens’ views of, and experiences with, service

delivery and governance
Assessing how service delivery of unions and upazila
parishads vary across districts
Determining the level of awareness of citizens and local
government officials on their mutual rights and
responsibilities;
Map 1: Spatial distribution of seven
intervention districts
Map 2: Contiguous districts of seven intervention
districts
Figure 1: Multistage Cluster sampling for Intervention
and Control Areas
Figure 1 shows sampling strategy for the baseline
study. As said before, the study team will consider
multistage cluster sampling to draw sample
households. Figure shows that 14 upazila clusters
will be selected. From 14 upazilas, 53 union
clusters out of 220 unions will be identified to take
1596 sampled households. In order to select
control districts, seven districts out of 27 districts
will be identified by using SEIFA index in two
levels. Then, 14 upazila clusters will be found from
223 possible upazilas. Later, 53 union clusters will
be selected from 224 possible union clusters. From
these 53 clusters, 1596 households will be selected
for survey.
Quasi-Experimental Design
• A number of writers have drawn attention to the possibilities offered by quasi-
experiments—that is, studies that have certain characteristics of experimental designs
but that do not fulfill all the internal validity requirements.

• Like a true experiment, a quasi-experimental design aims to establish a cause-and-

effect relationship between an independent and dependent variable.

• However, unlike a true experiment, a quasi-experiment does not rely on random

assignment. Instead, subjects are assigned to groups based on non-random criteria.

• Quasi-experimental design is a useful tool in situations where true experiments cannot

be used for ethical or practical reasons.

• A particularly interesting form of quasi-experiment occurs in the case of ‘natural

experiments’. These are ‘experiments’ in the sense of entailing manipulation of a social
setting, but as part of a naturally occurring attempt to alter social arrangements. In
such circumstances, it is invariably not possible to assign subjects randomly to
experimental and control groups.
Example of a true experiment vs a quasi-experiment
Example: True experimental design
To run a true experiment, you randomly assign half the patients in a mental health clinic
to receive the new treatment. The other half—the control group—receives the standard
course of treatment for depression.
Every few months, patients fill out a sheet describing their symptoms to see if the new
treatment produces significantly better (or worse) effects than the standard one.
However, for ethical reasons, the directors of the mental health clinic may not give you
permission to randomly assign their patients to treatments. In this case, you cannot run a
true experiment.
Example: Quasi-experimental design
You discover that a few of the psychotherapists in the clinic have decided to try out the
new therapy, while others who treat similar patients have chosen to stick with the normal
protocol. You can use these pre-existing groups to study the symptom progression of the
patients treated with the new therapy versus those receiving the standard course of
treatment. Although the groups were not randomly assigned, if you properly account for
any systematic differences between them, you can be reasonably confident any
differences must arise from the treatment and not other confounding variables.
Cross-Sectional Study
• A cross-sectional study is a type of research design in which you collect data from
many different individuals at a single point in time. In cross-sectional research,
you observe variables without influencing them.

• Researchers in economics, psychology, medicine, epidemiology, and the other

social sciences all make use of cross-sectional studies in their work. For example,
epidemiologists who are interested in the current prevalence of a disease in a
certain subset of the population might use a cross-sectional design to gather and
analyze the relevant data.

• The cross-sectional design is often called a survey design, but the idea of the
survey is so closely connected in most people’s minds with questionnaires and
structured interviewing that the more generic-sounding term cross-sectional
design is preferable. While the research methods associated with surveys are
certainly frequently employed within the context of cross-sectional research, so
too are many other research methods, including structured observation, content
analysis, official statistics, and diaries.
Cross-sectional vs longitudinal studies
Cross-sectional vs longitudinal studies: Example
You want to study the impact that a low-carb diet has on diabetes.
You first conduct a cross-sectional study with a sample of diabetes
patients to see if there are differences in health outcomes like
weight or blood sugar in those who follow a low-carb diet. You
discover that the diet correlates with weight loss in younger
patients, but not older ones.

You then decide to design a longitudinal study to further examine

this link in younger patients. Without first conducting the cross-
sectional study, you would not have known to focus on younger
patients in particular.
When to use a cross-sectional design
When you want to examine the prevalence of some outcome at a certain moment
in time, a cross-sectional study is the best choice.
Example
• You want to know how many families with children in New York City are
currently low-income so you can estimate how much money is required to
fund a free lunch program in public schools. Because all you need to know
is the current number of low-income families, a cross-sectional study
should provide you with all the data you require.
• Sometimes a cross-sectional study is the best choice for practical reasons
– for instance, if you only have the time or money to collect cross-sectional
data, or if the only data you can find to answer your research question was
gathered at a single point in time.
• As cross-sectional studies are cheaper and less time-consuming than
many other types of study, they allow you to easily collect data that can be
used as a basis for further research.
Descriptive vs analytical studies
Cross-sectional studies can be used for both analytical and descriptive
purposes:

• An analytical study tries to answer how or why a certain outcome might

occur.
• A descriptive study only summarizes said outcome using descriptive
statistics.

Descriptive vs analytical example

You are studying child obesity. A descriptive study might look at the prevalence
of obesity in children, while an analytical study might examine exercise and
food habits in addition to obesity levels to explain why some children are much
more likely to be obese than others.
How to perform a cross-sectional study
• To implement a cross-sectional study, you can rely on data assembled by another
source or collect your own. Governments often make cross-sectional datasets
freely available online.

• Prominent examples include the censuses of several countries like the US or

France, which survey a cross-sectional snapshot of the country’s residents on
important measures. International organizations like the World Health Organization
or the World Bank also provide access to cross-sectional datasets on their
websites.

• However, these datasets are often aggregated to a regional level, which may
prevent the investigation of certain research questions. You will also be restricted
to whichever variables the original researchers decided to study.

• If you want to choose the variables in your study and analyze your data on an
individual level, you can collect your own data using research methods such as
surveys. It’s important to carefully design your questions and choose your sample.
Cross-sectional research designs
More than one case. Researchers employing a crosssectional design
are interested in variation. That variation can be in respect of people,
families, organizations, nation states, or whatever. Variation can be
established only when more than one case is being examined.
Usually, researchers employing this design will select a lot more than
two cases for a variety of reasons: they are more likely to encounter
variation in all the variables in which they are interested; they can
make finer distinctions between cases; and the requirements of
sampling procedure are likely to necessitate larger numbers.

 At a single point in time. In cross-sectional design research, data

on the variables of interest are collected more or less simultaneously.
When an individual completes a questionnaire, which may contain
fifty or more variables, the answers are supplied at essentially the
same time. This contrasts with an experimental design. Thus, in the
classical experimental design, someone in the experimental group is
pre-tested, then exposed to the experimental treatment, and then
post-tested. Days, weeks, months, or even years may separate the
different phases
Cross-sectional research designs
 Quantitative or quantifiable data. In order to
establish variation between cases (and then to
examine associations between variables), it is
necessary to have a systematic and standardized
method for gauging variation. One of the most
important advantages of quantification is that it
provides the researcher with a consistent benchmark.

 Patterns of association. With a cross-sectional

design it is possible to examine relationships only
between variables. There is no time ordering to the
variables, because the data on them are collected
more or less simultaneously, and the researcher does
not (invariably because he or she cannot) manipulate
any of the variables. If the researcher discovers a
relationship between two variables, he or she cannot
What is survey research?

Survey research comprises a cross-sectional

design in relation to which data are
collected predominantly by questionnaire or
by structured interview on more than one
case (usually quite a lot more than one) and
at a single point in time in order to collect a
body of quantitative or quantifiable data in
connection with two or more variables
(usually many more than two), which are
then examined to detect patterns of
association.
 Longitudinal design(s)
The longitudinal design represents a distinct form of
research design. Because of the time and cost
involved, it is a relatively little-used design in social
research, so it is not proposed to allocate a great deal
of space to it. In the form in which it is typically found
in social science subjects such as sociology, social
policy, and human geography, it is usually an
extension of survey research based on a self-
completion questionnaire or structured interview
research within a cross-sectional design.

Consequently, in terms of reliability, replication, and

validity, the longitudinal design is little different from
cross-sectional research. However, a longitudinal
design can allow some insight into the time order of
 Longitudinal design(s)
With a longitudinal design a sample is surveyed and is
surveyed again on at least one further occasion. It is common to
distinguish two types of longitudinal design: the panel study and
the cohort study.

With the former type, a sample, often a randomly selected

national one, is the focus of data collection on at least two (and
often more) occasions. Data may be collected from different
types of case within a panel study framework: people,
households organizations, schools, and so on.

In a cohort study, either an entire cohort of people or a random

sample of them is selected as the focus of data collection. The
cohort is made up of people who share a certain characteristic,
such as all being born in the same week or all having a certain
experience, such as being unemployed or getting married on a
certain day or in the same week. The National Child
Development Study (NCDS) is an example of a cohort study.
Criteria in social research
Three of the most prominent criteria for the
evaluation of social research are
reliability, replication, and validity.
The suggestion that such criteria are mainly
relevant to quantitative research is
examined, along with the proposition that an
alternative set of criteria should be employed
in relation to qualitative research. This
alternative set of criteria, which is concerned
with the issue of trustworthiness
 Criteria in social research
Reliability
Reliability is concerned with the question of
whether the results of a study are repeatable. The
term is commonly used in relation to the question
of whether the measures that are devised for
concepts in the social sciences (such as poverty,
racial prejudice, deskilling, religious orthodoxy)
are consistent.
The quantitative researcher is likely to be
concerned with the question of whether a measure
is stable or not. After all, if we found that IQ tests,
which were designed as measures of intelligence,
were found to fluctuate, so that people’s IQ scores
were often wildly different when administered on
two or more occasions, we would be concerned
 Criteria in social research
Replication
The idea of reliability is very close to another
criterion of research—replication and more
especially replicability.
It sometimes happens that researchers choose
to replicate the findings of others. There may be
a host of different reasons for doing so, such as
a feeling that the original results do not match
other evidence that is relevant to the domain in
question.

In order for replication to take place, a study

must be capable of replication—it must be
replicable. This is a very obvious point: if a
researcher does not spell out his or her
Validity
 Validity: “the approximate truth of an inference [or
knowledge claim]” (Shadish et al., 2002: 34)

 Validity is an attribute of research findings, claims,

conclusions not of the research process

 There is no such thing as a perfectly valid claim 

research design can be used to address not solve
threats to validity.

 Validity refers to how accurately a method measures

what it is intended to measure. If research has high
validity, that means it produces results that
correspond to real properties, characteristics, and
variations in the physical or social world.
Validity and research design

problem simp
lified
rese
questions and arch
des
hypotheses ign

existing knowledge
methods and analysis
findings and conclusions
threats to validity

validity

43
 Typology of validity: strict vs. broader interpretations
‘Narrow’ interpretation (Shadish et al., Broader interpretation (Shadish et
2002) al., 2002)
Internal validity: Validity of inferences about Is there a causal relationship between
whether observed covariation between A (the two or more variables (e.g. a policy and
presumed treatment) and B (the presumed a target variable)?
outcome) reflects a causal relationship from A to
B as those variables were manipulated or
measured

Statistical conclusion validity: Validity of Is there a ‘true’ association between

inferences about the correlation (covariation) two variables? Does our estimate of a
between treatment and outcome parameter reflect the ‘true’ value?

Construct validity: Validity of inferences about Does a set of variables, indicators,

the higher order constructs that represent impressions (etc.) adequately reflect
sampling particulars (representation) the label, the concept, the
phenomenon that is being referred to?
External validity: Validity of inferences about To what extent can we generalize about
whether the cause-effect relationships holds over findings to other settings
variation in persons, settings, treatment (interventions, regions, target groups,
variables, and measurement variables etc.)?
(extrapolation) 44
Internal validity
 Conditions for causality in the strict sense, A causes B if:
1. A precedes B
2. A and B covary
3. We can distinguish the effect of A on B from other factors
affecting B

 Crucial dimension in experimentation; not necessarily a key

priority in other types of research

 Threats to internal validity  possible reasons why a

relationship is not causal: selection bias, other factors
affecting B

 Strengthening internal validity: use of (quasi-)experiments,

statistical controls, use of a theory of individual, social,
institutional behavior, look for falsifications
45
Statistical conclusion validity
 The validity with which statements about the association
of two variables can be made based on statistical tests.
 Addresses the question: Is there a relationship between
the two variables?

Threats to Statistical Conclusion Validity

Low statistical power: Is a threat when sample sizes are too
small or when alpha is set low. This is because low
statistical power increases the likelihood of making a
Type II error (accepting null when it is false).
Violated assumptions of statistical tests: Is a threat when
the assumptions underlying statistical tests (e.g.,
normality) are not met. This is because some statistical
tests cannot be meaningfully interpreted when
assumptions are violated.

46
Statistical conclusion validity
Fishing and error rate problem: Is a threat when
researchers make numerous multiple comparisons (when
conducting a large number of statistical tests). This is
because the likelihood of making a Type I error (rejecting
the null when it is true) increases the more comparisons a
researcher makes.

Reliability of measures: Is a threat when measures of

reliability for a scale are low. This is because an
unreliable scale cannot be relied on for detecting true
differences or changes

Random irrelevancies in the experimental setting: Is a

threat when irrelevant non treatment features of the
experimental setting influence scores on the dependent
variable

47
Construct validity
Do we adequately ‘cover’ a
phenomenon/object/concept in order to
generalize about it?

Are we using the right measures to ‘capture’

the concept?

Are there different interpretations about the

concept? If yes, how do we choose? Constructs
have social meaning!

Two levels: the concept/object/phenomenon

and the observation (respondent/case)  latter
case analogy with measurement validity 48
External validity
To what extent can we generalize about
findings to other settings (interventions,
regions, target groups, etc.)?
Generalization within and beyond the study
situation (sample)  three levels: from sample
to population (e.g. statistical inference from
random sample to population), from sample to
smaller groups or individual cases, from
sample to similar (similar or different)
samples/groups/settings
Threats to external validity: group/respondent-
specific, context-specific, intervention-specific
(etc.) features which limit generalizability
49
Thank you
Contact us.
Dr. Md. Zakir Hossain
Professor, URP Discipline, Khulna University
email: zakir@[Link]

50