Research and Sampling Design

 A research design is a plan, structure and strategy of

investigation conceived as to obtain answers to research
questions or problems.
 It is the complete scheme or program of the research.
 It includes an outline of what the investigator will do from
writing the hypotheses and their implications to the final
analysis of data
 A traditional research design is a blue print or detailed
plan for how a research study is to be completed.
 It includes choice and the measurement of variables,
selecting a sample, collection of data to be used as a basis
for testing hypotheses, and analyzing the results.
 Components of Research design:

1. Define the information needed

2. Design the phases of the research
3. Identifying variables & specifying the measurement and scaling
procedures
4. Construct & pretest a questionnaire (schedule) or an appropriate
form for data collection
5. Specify sampling process & sample size
6. Develop a plan of data analysis
7. Fixing time-schedule
8. Determine the source of fund and mechanisms to secure it
 A good research design takes in to consideration the
following factors.
1-The means of obtaining information
2-The availability and skill of researcher
3-The objectives of the problem to be studied.
4-The availability of time and money for the research
 In social sciences research the following two designs are
available.
◦ 1- Quantitative Research Design
◦ 2-Qualitative Research Design
◦ 3- Mixed Method
5.3.1. Quantitative Research Design
 Quantitative research design can be construed as a research

design that emphasizes quantification in the collection and

analysis of data.
 The measurement and classification requirements of the

information that is gathered under quantitative research

demands that the research designs are more structured,
rigid, fixed and predetermined to ensure accuracy.
 Quantitative research design focuses on objective and
standardized means of inquiry and application of statistical
analysis for attainment of objectivity and generalizations.
 In quantitative research design, the researcher identifies the
various components of the problem and operationalizes the
concepts into variables.
 Quantitative researchers mostly use questionnaire as major
methods to collect data.
 Standardized measures that fit diverse opinions and
experiences into predetermined response categories are used.
 This design facilitates to measure the reactions of a large
number of individuals to a limited set of questions which
facilitate easier statistical analysis.
5.3.2. Qualitative Design
 In some situations it becomes difficult to analyze a

phenomenon into various components or variables which

can be quantified.
◦ In such cases, the researcher takes into consideration the
phenomenon as a whole.
 Qualitative approach helps us to examine the nature of
human behavior and experience and social condition.

 It also permits the researcher to study selected issues, cases

or events in-depth.
 Qualitative research design can be construed as a research
design that usually emphasizes words rather than
quantification in the collection and analysis of data.
 It predominantly emphasizes on inductive approach to the
relationship between theory and research.
 Qualitative approach emphasizes on the ways in which
individuals interpret their social world .
 Qualitative research strategy embodies a view of social
reality.
 The main focus in qualitative research is to understand,
explain, explore, discover and clarify situations, feelings,
perceptions, attitudes, values, beliefs and experiences of a
group of people.
5.3.3. Mixed Methods
 Study designs in qualitative research are more appropriate

for exploring the variation and diversity in any aspect of

social life, whereas in quantitative research, they are more
suited to finding out the extent of this variation and
diversity
 For good quantitative research, it is important that you

combine quantitative skills with qualitative ones when

ascertaining the nature and extent of diversity and variation
in a phenomenon.
 This combining of qualitative research designs with

quantitative ones is what is understood as mixed methods.

 Rationale for conducting mixed methods may differ from
one research to the other
 Three of these approaches, ( triangulation, offset and
completeness) are briefly explained.
◦ Triangulation: implies that the results of an investigation
employing a method associated with one research strategy are
crosschecked against the results of using a method associated with
another strategy.
◦ Offset: implies that the weaknesses of a quantitative or a
qualitative method can be offset by including a qualitative or a
quantitative method that has its own strengths .
◦ Completeness: indicates that a more complete answer to a
research question or set of research questions can be achieved by
including both quantitative and qualitative methods. It implies that
the gaps left by one method (for example, a quantitative one) can
be filled by another (for example, a qualitative one).
 Subset of a larger population
 Sample
 Population or universe
 Population element
 Census
 Any complete group
◦ People
◦ Sales territories
◦ Stores
 Investigation of all individual elements that
make up a population
Stages in the Define the target population
Selection
of a Sample Select a sampling frame

Determine if a probability or nonprobability

sampling method will be chosen

Plan procedure
for selecting sampling units

Determine sample size

Select actual sampling units

Conduct fieldwork
 Relevant population
 Operationally define
 A list of elements from which the sample may
be drawn
 Working population
 Mailing lists - data base marketers
 Sampling frame error
 Group selected for the sample
◦ Primary Sampling Units (PSU)
◦ Secondary Sampling Units
◦ Tertiary Sampling Units
 The difference between the sample results
and the result of a census conducted using
identical procedures
 Statistical fluctuation due to chance variations
 Nonsampling errors
 Unrepresentative sample results
 Not due to chance
 Due to study design or imperfections in

execution
 Probability sampling
 Known, nonzero probability for every
element

 Nonprobability sampling
 Probability of selecting any particular
member is unknown
 Convenience
 Judgment
 Quota
 Snowball
 Also called haphazard or accidental sampling
 The sampling procedure of obtaining the

people or units that are most conveniently

available
 Also called purposive sampling
 An experienced individual selects the sample

based on his or her judgment about some

appropriate characteristics required of the
sample member
 Ensures that the various subgroups in a
population are represented on pertinent
sample characteristics
 To the exact extent that the investigators

desire
 It should not be confused with stratified

sampling.
 A variety of procedures
 Initial respondents are selected by probability

methods
 Additional respondents are obtained from

information provided by the initial

respondents
 Simple random sample
 Systematic sample
 Stratified sample
 Cluster sample
 Multistage area sample
 A sampling procedure that ensures that each
element in the population will have an equal
chance of being included in the sample
 A simple process
 Every nth name from the list will be drawn
 Probability sample
 Subsamples are drawn within different strata
 Each stratum is more or less equal on some

characteristic
 Do not confuse with quota sample
 The purpose of cluster sampling is to sample
economically while retaining the
characteristics of a probability sample.
 The primary sampling unit is no longer the

individual element in the population

 The primary sampling unit is a larger cluster

of elements located in proximity to one

another
Population Element Possible Clusters in Ethiopia

Ethiopian farmers Regions

Zones
Woreda
Kebele
Teams
Households
 Degree of accuracy
 Resources
 Time
 Advanced knowledge of the population
 National versus local
 Need for statistical analysis
 Internet surveys allow researchers to rapidly
reach a large sample.
 Speed is both an advantage and a

disadvantage.
 Sample size requirements can be met

overnight or almost instantaneously.

 Survey should be kept open long enough so

all sample units can participate.

 Major disadvantage
◦ lack of computer ownership and Internet access
among certain segments of the population
 Yet Internet samples may be representative
of a target populations.
◦ target population - visitors to a particular Web
site.
 Hard to reach subjects may participate
 Unrestricted samples are clearly convenience
samples
 Randomly selecting visitors
 Questionnaire request randomly "pops up"
 Over- representing the more frequent visitors
 Sample size is a function of:
• size of the population of interest,
• the desired confidence level, and
• level of precision.
 You can calculate a formula to determine the
appropriate sample size or you can use tables, a Guide
to Minimum Sample Size.
• Note that these are minimum sample sizes. Whenever possible,
you should select a larger sample size to compensate for the
likelihood of a lower than 100% response rate.

38
 Confidence level: How certain you are (or need to be)
that the statistic obtained from your sample is an
accurate estimate of the population as a whole.
• The issue of how much sampling error can be tolerated only
applies to probability samples.
• Confidence level refers to the percentage that the sample
differs from the theoretical population within the confidence
interval.
• It can range from one percent to 99 percent.
• Realistically, confidence level should never fall below 90, 95
and 99 percent being the most common.

39
 Confidence interval: The calculated range within
which the true population value lies and for which we
can express 95% confidence level(the standard, but
may vary).
• So if the significance interval is three percent, it would mean that the
data from the sample fall within the three percent range as compared
to what the results would be using the whole theoretical population.

• More concretely, if President X’s approval rating is 61 percent according

to the latest opinion poll, the President’s rating is actually between 58 and
64 percent among all adults.

40
 For example: you might read in the paper that 48% favor
raising taxes and 52% oppose raising taxes .
 Assuming Significance level is (+/-5%)
 What this means is that if everyone in the population were
asked, the actual proportions would be somewhere between
43% to 53% (48 +/- 5) favoring raising taxes, and 47% to 57%
(52 +/- 5) _opposing.
 Most evaluations accept a sampling error of 5%.
 In the tax example, if we had a 5% margin of error, then the
true picture of opinions would be between 43% to 53% favoring
raising taxes and between 47% to 57% opposing raising taxes.

41
Option 1: For dichotomous/binomial variable or for population
characteristic expressed as a proportion (Approve or disapprove, for
or against, yes/no)

n= p% x q% x [z/e%]2
Where ,
◦ n is the minimum sample size required ,
◦ P% is the proportion belonging to specified category
◦ q% is the proportion not belonging to the specified category
◦ z is the z value corresponding to the level of confidence required
◦ e is the margin of error required
Level of confidence Z value
90% certain 1.65
95% certain 1.96
99% certain 2.57
42
 Assume that 30 percent of the students say they will support the current
continuous assessment approach. We decide to estimate the true
proportion in the population with in 10 percentage points of this figure
(p= 30% + or -). Assume further that we want to be 95 percent confident
that the population parameter is within plus or minus of the sample
proportion.
 We can be 95 percent confident that 30 percent of the respondents
would say they would support the approach with a margin of error of +
or- 10
Find the minimum sample size?
n= p% x q% x [z/e%]2
 P% =30% Or Given: P= 0.3, q=0.7, z = 1.96, e=0.1
 q% =70%
 z =1.96
 e = 10%

n= 81
43
Exercise:
 To answer a research question you need to estimate the
proportion of total population of 4000 home care clients who
receive a visit from their home health care (extension workers)
at least once a week. You have been told that you need to be 95
percent certain that the estimate is accurate ( the level of
confidence in the estimate); this corresponds to a z score of
1.96. You have also been told that your estimate needs to be
accurate with in plus or minus 5 percent of the true percentage
(the margin of error that can be tolerated).
 You still need to estimate the proportion of responses who
receive a visit from their home health care assistant at least once
a week. From your pilot survey you discovered that 12 out of
the 30 clients receive a visit at least once a week.

Question: Calculate the minimum sample size

44
 Solution:
Given:
12/30= 40%---- P%= 40%
q% =60%
z= 1.96
e= 5%
n= p% x q% x [z/e%]2

n= 369

45
 Let us assume that a political party wants to conduct a poll to
estimate the percentage of voting for the party within ±5 per
cent points and that the party wishes to be 95 per cent confident
of the result. Also assume that the percentage of voting for the
party is believed to be 35 per cent. What sample is appropriate
in this case?

◦ Solutions: We have Z = 1.96, p = 0.35, d = 0.05

◦ n= 350

46
 • Determine the size of the population.
• Decide on the margin of error. (As much as possible the margin of
error should not be higher than 5%. Probably 3% is an ideal one)
• Determine sample size

n= N__
1+NE²
n = sample size
N = the size of the population
e = the margin of error

 Compute the sample proportion by dividing the sample

obtained by the population.
47
• Population is 5,346
• Margin of error is 3%
• Using the formula: n = N__
1+NE²
n= 5346__
1+ 5346(.03) ²
n = 920

48
o Though we use the above formulas, there are no fixed rules in
determining the size of a sample needed.

o There are guidelines that should be observed in determining

the size of a sample.
o When the population is more or less homogeneous and only the
typical, normal, or average is desired to be known, a smaller
sample is enough. However, if differences are desired to be
known, a larger sample is needed.

o When the population is more or less heterogeneous and only the

typical, normal or average is desired to be known a larger sample
is needed. However, if only their differences are desired to be
known, a smaller sample is sufficient.

49
 Refer to the standardized table (that can be
obtained from research methods’ books) to
find minimum/base sample size

50
End of Chapter Five