2. Outline
Research Design
What is Research Design?
Types of Research Designs
Sampling Design
– Why sampling?
– What is a good sample
– Types of sampling: Probability vs Non-probability
3. Research Design
A research design constitutes the blueprint for
the collection, measurement and analysis of
data.
Research design could be classified in different
ways based on:
1. the degree to which the research question has
been crystallized
2. time dimension
3. the purpose of the study
4. the research environment
5. the source of data and analytical method
4. Research Design Cont...
1. Based on the degree to which the research
question has been formulated
• Exploratory: has loose structure with the objective of
discovering future research tasks
• Formal study: involves precise procedure & data source
specification and the aim is to test hypothesis or answer
research question
• The exploratory-formal study dichotomy is less precise
classification as all studies have some elements of
explorations
• Two stage design is possible. First exploratory then followed
by causal/more formal in this case.
2. Based on time dimension
• Cross-sectional studies are carried out in a snapshot.
• Longitudinal studies require repeated measurements. It
gives the possibility of capturing changes of variables over
time.
5. Cont…
Based on the purpose of the study
• Exploratory study
• Descriptive study when the objective is to find out who, what,
where, or how much.
• Correlational study
• Evaluative study
• Causal/Explanatory if the aim is to find out why something is
happening or how one variable produces changes in another it is
causal.
Based on the research environment
• Statistical study:
– Designed for breadth rather than for depth; qualitative study is for
depth.
– It attempts to capture xics of a population from the xics of the sample.
– It tests hypothesis quantitatively.
– Generalizations about findings are presented based on the
representativeness of the sample.
6. Cont…
• Case study:
– It emphasizes on a full contextual analysis of fewer
events/cases and their inter-relations.
– It relies much on qualitative data, and hence makes
rejection or support of hypothesis more difficult, if any.
– It seeks for detail information from multiple sources.
– Hence, provides valuable insight for problem solving,
evaluation & strategy. It allows for verifying evidences
and avoids missing data.
Based on the source of data and analytical method
• Laboratory research
• Field research
• Simulation
7. Cont…
Generally, we have three types of research designs
1. Exploratory study:
• Conducted to clarify ambiguous situations or discover ideas
that may be potential business opportunities. Exploratory
research is particularly useful for new product/service
development
• It is not intended to provide conclusive evidence from which
to determine a particular course of action.
• The objective is to generate more information about a
situation before launching a formal study
• It is done to define clearly the research question in the form
of investigative questions
• Helps to identify (extraneous) variables that can be ignored
• Suggests if doing additional and more formal research is
feasible
• Although both qualitative & quantitative methods could be
used, it relies heavily on qualitative methods
8. Cont…
The following approaches could be used in
exploratory research:
– In-depth interviewing, which is not structured
– Participant observation
– Photographs and videotaping
– Role play
– Case studies for an in-depth contextual analysis
– Key informants/elites interviewing
– Documents analysis
9. Cont…
2. Descriptive study:
• Used to describe characteristics of objects, people, groups,
organizations, or environments.
• More structured in terms of stating research questions
(hypothesis).
• It serves to achieve a variety of research objectives:
– To describe the characteristics of the study subject (who, what,
when and how)
– To estimate the proportions of a population that have particular
characteristics
– To discover association/correlation among different variables
Example:
In a saving association one might be interested to develop the profile
of savers such as: Age, sex, amount saved, the number of accounts
opened within the last six months, frequency of withdrawal per year,
distance of the individual from the main office, education level,
family size, etc
10. Cont…
• It could be simple or complex. The simplest form
of descriptive study addresses only a uni-variate
question or hypothesis to state the size, form,
distribution or existence of a variable.
• In its complex form, it demands to collect
information about multiple variables and carry
out chi-square/cross-tabulation analysis and
correlation matrix analysis
Example:
• Association/correlation could be done between
amount saved and income/family size
11. Cont…
3. Causal study
• It allows to make causal inferences
• It seeks to identify cause-and-effect relationships.
• Correlation is different from causation.
• In causation, “A” forces “B” to occur or “A” is
responsible for the changes occurred in “B”.
• In testing causal hypothesis, we collect evidence
that increases our belief that A leads to B.
• Criteria for causality:
– Concomitant variation
– Temporal sequence
– Non-spurious association
12. Cont...
1. Concomitant variation: Is there a predicted co-variation between
“A” & “B”?
• When “A” doesn’t occur is there also an absence of “B”?
• When there is less of “A”, do you find more/less of “B”?
2. Temporal sequence: Is there time order of events moving towards
the hypothesized direction?
• Does “A” occurs before “B”? Or the cause must occur before the
effect.
3. Non-spurious association: is it possible to eliminate other possible
causes of B? Is the co-variation between a cause and an effect
is true, rather than due to some other variable?
Theory helps to rule out spurious association
Often, a causal inference cannot be made even though
the other two conditions exist because both the cause and
effect have some common cause.
– Eg. Correlation of ice cream consumption & crime rate:
Inflation could be a cause for both
13. Cont…
Degree of causality
• Three types of causality based on degree
1. Absolute causality: the cause is necessary & sufficient
to bring the effect
– In behavioral science it is a rarity
2. Conditional causality: the cause is necessary but not
sufficient to bring the effect
– The cause can bring about the effect but it can not do so
alone
– Smoking is a conditional cause for cancer as there are
some other variables disposing oneself to cancer
3. Contributory causality: the cause is neither necessary
nor sufficient to bring the effect
How do you measure degree of causality in research?
R2 measures degree of causality
14. Comparisons
Criteria Exploratory Descriptive Causal
Amount of uncertainty for
decision making
Highly ambiguous Partially defined Clearly defined
Key research statement Research question Research question Research
hypothesis
When conducted? Early stage of decision
making
Later stage of
decision making
Later stage of
decision making
Research approach Unstructured Structured Highly structured
Nature of result Discovery oriented often
in need of further
research
Result can be
managerially
applicable but still
further research
may be needed
Confirmatory
oriented & fairly
conclusive with
managerially
actionable results
15. Cont…
• The different types of research designs
discussed here serve each other as building
blocks
– Exploratory research builds the foundation for
descriptive research,
– Descriptive research usually establishes the basis
for causal research.
16. Sampling Design
Samples and Sampling:
Samples: are subset or portion of a population
Sampling: is the procedure of drawing a portion of the
population.
Sample statistics: estimates of xics of a population based on a
sample.
Population: Any complete group of entities that share some
common set of characteristics.
Census: an investigation of all the individual elements that make up a
population.
Parameter: estimates of population xics based on census.
Remark: the purpose of sampling is to estimate the population
parameter from the sample statistics. The results of a good sample
should have then the same characteristics as the population as a
whole, i.e sample statistics should be closer to parameters.
17. Why sampling?
Do we always do census? If not why sampling?
– Time and budget constraint
• Working on samples cuts down substantial amount of
money and time
– Samples yield accurate & reliable results
• Interviewer mistakes, tabulation errors, and other non-
sampling errors may increase during a census because of the
increased volume of work.
• In a sample, increased accuracy may sometimes be possible
because the fieldwork and tabulation of data can be more
closely supervised.
– Destruction of test of units
• If the nature of the study involves destructing the study
units/subjects, doing census destroys the entire population.
18. Types of sampling design
• There are two types: probability & non-
probability sampling design
• Probability sampling design includes:
– Simple random sampling
– Systematic sampling
– Stratified sampling
– Cluster sampling
– Multi-stage sampling
• Non-probability sampling design includes:
– Convenience sampling
– Purposive sampling
• Judgmental sampling
• Quota sampling
– Snowball sampling
19. Cont…
• Probability sampling: each member of the
population has a known, non-zero chance of
being selected, i.e, Random Selection.
• Non-probability sampling: selection of samples is
non-random and subjective or based on personal
judgment, i.e, the probability for a study unit
being chosen is unknown.
• Issues in sampling design:
– Decide on the unit of analysis
– Determine/Identify the relevant (target) population
– Identify the sampling frame if probability sampling
design
– Decide the type of design
20. Cont…
• Determining your population:
– This might be apparent from the management
problem/objective
– Decide on your unit of analysis: individuals,
households, companies, etc
• Sampling frame: the list of elements from
which the sample is to be drawn. It is closely
related to the population but in practice
sampling frame often differs from theoretical
population
• Type of sampling design:
– Depends on the objective and nature of study
21. Probability sampling design
1. Simple random sampling technique
– When the population is relatively homogenous for issue
we studying
– Each member of the population has equal chance of
being selected
– Use lottery method/random number tables to draw a
sample from the sampling frame
2. Systematic (random) sampling
– Arrange the sampling frame in some order (eg.
Alphabetical)
– Decide the sample size (n)
– Divide the total population (N) with the sample size (n),
i.e, N/n to arrive at the interval for drawing sample
– Draw the N/nth from the sampling frame until you finish
drawing “n” size of the population
22. Cont…
3. Stratified sampling
– When we have variation in the population on the
parameter we are studying/measuring
– We stratify/group the population using criteria such as
sex, age, education, batch, customer status, etc
– Within stratum there is homogeneity but between
strata there is heterogeneity
– Decide the sample size, and decide how much sample
to take from each stratum
– This could be done based on proportion to the size of
the stratum
– Employ either simple random sampling or systematic
random sampling technique to draw samples from
each stratum
23. Cont…
4. Cluster sampling
– Geographical regions are commonly considered as
clusters
– Heterogeneity within cluster but homogeneity among
clusters
– Random sampling of clusters then interviewing entire
members of the cluster
– Used when sampling frame is not available
– Economically efficient sampling technique instead of
travelling all the regions/geographically dispersed
areas
– The primary sampling unit is not the individual
element in the population but a large cluster of
elements;
– Eg. Consumer feedback on new products
24. Cont…
Multi-stage sampling technique
– This is a method of employing mix of different
sampling techniques
– For example, in the first stage you may classify the
population into certain strata (groups) based on some
stratification variable
– In the second stage, either you can employ simple
random sampling or systematic sampling technique
– You can do similarly in cluster sampling as well
– It is also possible to mix probability and non-
probability sampling designs.
– For example, in the first stage a researcher might
select a certain group in the population purposively
(non-probability sampling) but in the second stage
either simple random or systematic sampling
technique could be adopted
25. Comparisons of probability sampling techniques
Sampling
technique
Cost Degree of
use
Advantage Disadvantage
Simple random High cost Moderate Easy to analyze data and
sampling error
Requires sampling frame;
Large errors for same sample
size compared to stratified;
High cost if respondents are
dispersed
Systematic Moderate
to high
cost
Moderate Simple to draw sample Requires sampling frame
Stratified High cost Moderate Ensures group
representation; Comparison
among strata possible;
reduce variability
Requires information for
stratification variable;
Sampling frame needed
Cluster Low cost Frequent Lowers field cost; Possible
to estimate Xics of clusters
Larger errors for same sample
size compared to other
techniques
26. Non-probability sampling design
1. Convenience sampling
– The sampling procedure of obtaining those people or units
that are most conveniently available
– Economical and fastest way of getting questionnaire filled up
2. Purposive sampling: We select a particular group of units from the
population based on reason/purpose. We need to justify the
reason why we choose a particular unit
– Judgmental sampling: we use our judgment whether it is
appropriate to choose a particular unit from the population
– Quota sampling: we assign quota for group of units in a
population in drawing samples
3. Snowball sampling
– In some studies identifying the population is very difficult
– We try to find one study unit from the population and we build
our sample based on the recommendation of the first study
unit.
27. Comparisons of non-probability sampling techniques
SN Sampling
technique
Cost Degree of
use
Advantage Disadvantage
1 Convenience Very low Extensively
used
No need for list of
population
Samples are
unrepresentative;
Making inference beyond
sample risky
2 Judgmental Moderate cost Average Sample guaranteed to
meet a specific
objective; useful for
certain types of
forecasting
Bias due to expert’s
beliefs make samples
unrepresentative;
Making inference beyond
samples risky
3 Quota Moderate cost Very
extensively
used
Requires no need for
list of population
Introduces bias in
researcher’s
classification of subjects;
Making inference beyond
sample risky
4 Snowball Low cost Used in
special
situations
Useful in locating rare
population
High bias because
sample units are not
independent; Inference
beyond samples risky
28. Sample size
• How big should be the sample size?
– 10%, 15%, 30% or what?
• The size of the sample depends on:
– The variation in the population parameter of the
study, i.e the population variance. The greater the
dispersion/variance, the larger the sample must
be.
– The precision level or the magnitude of acceptable
errors
• Precision is measured by: the interval range to find the
parameter estimate and the degree of confidence
29. Cont..
–The confidence level: A percentage value that tells how confident a
researcher can be about being correct. It could be either 90%, 95%, or
99%. It tells you how sure you can be. It is expressed as a percentage
and represents how often the true percentage of the population who
would pick an answer lies within the confidence interval. For example,
95% confidence level means that if you had conducted the same
survey 100 times, 95 times out of 100 the survey would have yielded
the same results.
–A confidence interval estimate is based on the knowledge that
population mean is equal to sample mean ± a small sampling error. It
is the plus-or-minus figure usually reported in newspaper or television
opinion poll results. For example, if you use a confidence interval of 4
and 47% percent of your sample picks an answer you can be "sure"
that if you had asked the question of the entire relevant population
between 43% (47-4) and 51% (47+4) would have picked that answer.
–When you put the confidence level and the confidence interval
together, you can say that you are 95% sure that the true percentage
of the population is between 43% and 51%.
30. Cont…
Steps in estimating sample size:
1. Estimate the standard deviation of the
population
– From previous study or pilot study
2. Decide on the magnitude of error
3. Determine the confidence interval
4. Calculate sample size as follows:
n =(ZS/E)2 where,
n=Sample size; Z= standardized value;
S= Estimate of the population standard deviation;
E=Magnitude of error
31. Cont…
Standard error (Z) Percent of area Degree of confidence
1.00 68.27 68%
1.65 90.10 90%
1.96 95.00 95%
3.00 99.73 99%
32. Cont...
• It is possible to use one of them to construct a
table that suggests the optimal sample size –
given a population size, a specific margin of
error, and a desired confidence interval.
• This can help researchers avoid the formulas
altogether. The table below presents the
results of one set of these calculations. It may
be used to determine the appropriate sample
size for almost any study.
35. Exercise
• Suppose a survey researcher studying annual
expenditures on lipstick wishes to have a 95
percent confidence level (Z=1.96) and a range
of error (E) of less than $2. If the estimate of
the standard deviation is $29,
– Question 1: Calculate the sample size?
– Question 2: Calculate sample size if the allowable
error term is doubled?
36. Answer & Conclusion
• Q1: n=808
• Q2: n=202
• Doubling the range of acceptable error reduces
sample size to approximately one-quarter of its
original size.
• Stated conversely in a general sense, doubling
sample size will reduce error by only
approximately one-quarter.