5.

Research Methodology
• Comprises all scientific methods employed to
answer the RQs
• It includes all issues of research design, data,
sample, measurement and analysis
 A: Research Design and Procedures
 State your design and rationale for the choice
o What is the study about?
o Where the study is carried out?
o What type of data are used?
o What is the sample frame and how the sample is selected?
o What data collection methods?
o Unit of analysis
o How data are analyzed?
1) Sampling Design
Sampling: A procedure of selecting elements from a larger
group (population).
It is a process of using a small number of items of a larger
population to make conclusions about the whole population.

What is the difference between a Census and Sample survey?

Conducting census survey is very costly in terms of both money and

time.

It is therefore normally conducted by governments and in rare

cases.
 Sampling process
1. Define the population
2. Identify the sampling frame
3. Selecting a sampling procedure
4. Determine the sample size
5. Select the sample elements
6. Collect the data
 Define the population
• A population (Universe) is defined as the
totality of cases that conform to some
designated specification (criteria for
‘membership’).

• The population can be referred to as the level

of generalization (of the result of the study).
 Sampling frame
• Sampling frame is the listing of the elements
from which the actual sample will be drawn.
• This is also known as source list.

• Such a list should be representative,

comprehensive, correct and reliable.
e.g. A list of all manufacturing firms in TZ
 Sampling Techniques
There are mainly 2 types of sampling techniques
Non-probability sampling
• The sample is not based on random sampling, and the
results cannot be generalized to the population.

Probability sampling
• Each element/unit of the population (or, sampling frame)
has a known and equal chance of being included in the
sample
 Non-probability samples

a) Convenience sampling technique (or,

accidental sampling)

Selecting sample elements that are most

readily available to participate
b) Judgment sample (or, purposive
sample)
• The sample elements are chosen purposely just
because it is expected that they can serve the
research purpose.

• For example, selecting HR Manager from the firm.

c) Snowball sampling
• The initial respondents may be chosen using
a probability method
• The initial respondents are used to identify
(and inform) about other respondents in the
population
• This process continue until the required
sample size is reached.
 Probability sampling

a) Simple random sampling

• Random sampling: Is a sampling procedure

whereby each population element has an
equal chance of being included in the
sample.
b) Systematic sampling
• Is a sampling procedure in which an initial
starting point is selected by a random
process, and thereafter every nth number on
the list is selected.
c) Stratified sampling
• Is a sampling procedure in which simple random samples
are drawn from within different strata that are more or less
equal on some characteristics

• Divide the parent population into mutually exclusive and

exhaustive groups or subsets (strata).

• A simple random (or, systematic) sample is chosen

independently from each group or stratum.
d) Cluster sampling
• This procedure is similar to stratified sampling
technique except that cluster sampling involves
selection of the entire group.

• The parent population is divided into mutually

exclusive and exhaustive subsets which should be
formed to be as heterogeneous as possible.

• Select a random sample of the subsets

 Determining the sample size
Sample size depends on:
• Type of study (Qualitative versus Quantitative)
• Time & Money
• Number of categories, subgroups – use of statistical
methods
• Type of sampling method used
• The relative size n (sample size)/N (population size)
• The homogeneity of the population (standard
deviation)
Important statistical ‘laws’:

Taking a random sample of size n from a

population of size N, is statistically
equivalent with taking one sample of all
possible samples that can be drawn from the
parent sample.
The Central-Limit Theorem

If simple random samples of size n are drawn from a

population, then, when n is large, the sample mean
(of all possible samples) will be approximately
normal distributed with the mean equal to the
population mean, and variance equal to the
population variance.
How will you know the required sample size?
Data Collection Methods
Types of Data

Primary data: These are data collected afresh and for the first
time.

Secondary data: Data already collected by someone else.

• Types of data
—Cross section data
—Time series data
—Panel or longitudinal data
Methods of collecting primary data
—Observation
—Interview
—Questionnaire
Observation: Researcher’s direct observation of a phenomenon without
asking from the respondents

Interview: Can be personal interview, focus group interviews (5-10).

Questions are non structured and open.
Interviews can be conducted through face to face, telephone, etc

Questionnaire: Is a form of personal interviews but with structured questions

Questionnaires can be administered face to face, telephone, mail etc

Read Kothari, C.R (2004: pp.118-119) for guidelines on how to construct

questionnaires and interviews.
Methods of collecting secondary data
• Documentary Review (is not the same as literature
review!!!)
— Reports: e.g.; annual, semi annual and quarterly
reports
— Memos: e.g.; office memos

— Letter: e.g. official communication letter

— Internet source: e.g. Website of the National Bureau

of Statistic, WB, IMF
Measurement Scales
 Measurement: Is the process of assigning numbers to objects
or observations in order to state the degree of presence or
absence of a phenomenon.
 We measure physical objects as well as abstract concept

o Physically objects: Directly observable, e.g. people, firms,

price, weight, height etc
o Abstract concept: Can’t be observed directly, e.g. intelligent,
satisfaction
o Physical objects can easily be measured using standard unit of
measurement.
o Abstract concepts are difficult to measure and thus require
critical attention
 The measurement scales for both physical and
abstract concepts can be considered in terms of
their mathematical properties. The usual
classification of measurement scales are

1. Nominal scale
2. Ordinal scale
3. Interval scale
4. Ratio scale
Nominal scale: This is the simplest and lowest scale of
measurement.
 It a system of assigning numbers, symbols or letters
to objects or events in order to serve as labels.
 For example, numbers assigned to football or
basketball players. These numbers are for
identification only.
Goalkeeper = 1
Striker = 10
 Does it mean that player 10 plays more than player
1? ( 10 > 1) NO! NO! NO!
 Numbers are only for identification and does not give
any meaningful quantitative value

 Therefore, the only arithmetic operation with

nominal scale in the case of measures of central
tendency is MODE. No measures of dispersion are
used with nominal scale.

 In case of test for statistical significance, the only

statistical technique used is the Chi-square test (χ2).
In the case of test for relationship (association), the
contingency coefficient may be used.
 However, nominal data are very useful in classifying
objects into various categories.

 For example, categorizing respondents by

demographic characteristics; Gender (Male = 0
Female = 1); location (rural = 0 urban = 1) etc.
Ordinal scale: This scale arranges objects in the order
between lowest and highest but there is no any
attempt to report by how much difference because
the intervals between ranks are not equal.
 This scale is very useful when the researcher
examines respondent's preferences or satisfaction
about a certain phenomenon.
For example,
 How satisfied are you with our product/service?
1 excellent 2 good 3 fair 4 poor
 Can you rank our products based on this scale?
1 excellent 2 good 3 fair 4 poor

 The appropriate measure of the central tendency is

the MEDIAN while percentiles can be used as
measures of dispersion. Non-parametric statistic can
be used.
• Non-parametric data involves data which do
not conform to assumptions of normal
distribution, e.g.; ordinal data

• The appropriate statistical test for non-

normality data is χ2
Interval scale: This is the scale that arranges objects not only by
their order but also distinguishes this ordered arrangement in
units of equal intervals.

 The main limitation of interval scale is the lack of a true zero

because it has no ability to measure the absence of a
characteristic of a phenomenon. .

 But, you can not conclude that 60 degrees is as twice as

warmer as 30 degrees because zero does not represent lack
of temperature but a relative point on the temperature scale.
 Interval scale provides more powerful
measure than ordinal scale and most statistics
such as mean and standard deviation can be
computed.

 Correlation analysis can be analysed and test

for statistical significance such as
t-test and F-test can be used
Ratio scale:
 Ratio scale consists of a true zero of
measurement. For example, the zero-point on
a centimetre scale indicates the complete
absence of length. All statistical techniques
can be applied with ratio scale.

 Most measurements in physical science uses

ratio scale
Duration, energy, mass, weight etc