Research Methodology &

Statistics

AGEC 603
Lecture 6

Charles Yaw Okyere (Dr. Agr.)

 OUTLINE
QUESTIONNAIRE DESIGN
SAMPLING TECHNIQUES AND
ANALYZING SAMPLE DATA
TAKING A SAMPLE
Scientific method of enquiry—recall
Hypothesis – if (A) then (B)
Operationalization of concepts –
describe A/B to real world analogy
Collect data / facts on A/B
Subject hypothesis to validation
 Sources of Data for Research
Time Series data (High/low frequency)

Cross-sectional (usually micro-level) data

Combination: Panel data

Sources of these data differ: Micro-level data
collection deals with entire populations. Resource
constraints hence:
population → sample → sampling required
 Use of Questionnaires (1)
Use Questionnaires to collect data:
ask people for:

(a) quantitative data

Figures
Statistics
Amounts
 Use of Questionnaires (2)
ask people to/for:
(b) qualitative data

Describe conditions and procedures that affect their

work
Their judgment and views about processes,
performance, adequacy etc
Report past events and also to make forecasts
Tell about their attitudes and opinions
Describe their behavior and that of others
 Who do we ASK?
Because answers to questionnaires rely on
people to provide answers:

People with the ability to observe, select, acquire,

process, evaluate, interpret, store, retrieve and
report can be a valuable source of information
under the right circumstances!

However the human mind is a very complex and

a vulnerable observational instinct!
Advantages to use of questionnaires
Questionnaires allow:

To answer descriptive and normative or evaluative

questions
Cost effective way to collect a large amount of
standardized information
More versatile than other methods since it
involves people who can report facts, figures
amounts, dates, attitudes, opinions etc.
 Disadvantages
Difficult to use if respondent population cannot be
readily identified
If respondents are unwilling to provide accurate
information
If it taxes the limitations of respondents likely to
have inaccurate results, etc.
If not properly planned out, inferences become
problematic
 Methods of collecting the data (1)
(a) Quantitative:
1. Structured survey: entails administering a
written questionnaire to a sample of respondents.
Advantage: a standardized set of questions to all
that eliminate biases due to enumerator style.
2. May be obtained from the records/files of
projects/institutions/farms

Quantitative data necessary to estimate rates, a

number, proportion related to the target.
 Methods of collecting the data (2)
(b) Qualitative: 3 approaches

(1) In-depth interview with individual respondents.

Usually with “key informants” (village chief,
extensive worker, teachers etc)

The interviewer gently probes the respondent

which permits them to have a conversation in
which ideas flow freely.
Methods of collecting the data (3)
(2) Group interviews – community meetings or
focus group discussions (FGDs). Best conducted
by two or more interviewers who address queries
to the participants on the focused topic.
(3) Participant observation: the interviewer is
involved directly to have an extensive observation
of the activity, behavior or relationship.
Qualitative data are needed when the attitudes,
beliefs and perceptions of the target population must
be known in order to understand its reaction to some
service
Tasks in questionnaire framework
Plan questionnaire
Develop measures
Design the sample (to elaborate later)
Develop and test the questionnaire
Produce the questionnaire
Prepare and distribute (through mail/direct
application)
Collect the data (if it is by direct interview skip)
Collate, reduce and analyze data.
Planning the questionnaire
Developing the measure
Designing the sample
Designing the sample or population for
data collection
Survey population
Selecting the sample
 Sampling technique (sampling
survey methods)
Non – statistical sampling.
Statistical sampling.
 Sampling Techniques
Sampling
Sampling techniques provide a range of methods
that enable you to reduce the amount of data you
need to collect from a population by considering data
from only a sub-group rather than all possible cases
Sample
A smaller unit bearing the same characteristic (s) of
the population of interest. Researchers often collect
data from a sample to demonstrate that sample is
representative of population from which it is drawn.
 Sampling Techniques……..
The sampling process:

Definition of the population of concern

Specification of the sampling frame

Specification of the sampling method

Sampling and data collection

 Sampling Techniques……
Two Types of Sampling Techniques

Probability- representative sampling/statistical sampling

Each element in the population has an equal (non-zero) chance of
being selected. Generalization (inferential / inductive reasoning)
possible because of equal chance selection for sample inclusion
Non-probability- judgment sampling
Not all elements/cases have an equal chance of being included in
the sample
The results of non-probability sampling may be biased – the
sample results may not be representative of the population
Examples are convenience and judgment sampling
 Statistical Sampling
A sampling procedure by which every element in the population has a known and usually equal chance of being chosen for
inclusion in the sample.
For a sample to be random, this must be characterized by:
Clearly defined selection procedures.
The use of lists (or their equivalent) as sampling frames.
The applicability of sampling theory (assumption that every member of the sample has a known and a non – zero probability
of selection).
The possibility of estimating sampling error: population mean – sample mean ( ).

The statistical sample / random sample is also known as

Probability sampling (concepts of probability are applied to sample selection).
Scientific sampling (can be used as the basis for statistical inference).
 Probability Sampling Techniques

Simple random sample

Systematic sample
Stratified sample
Cluster sample
Exit polls: an implicit list
Others?
 Probability Sampling Techniques
Simple random sample: Selection based entirely on
chance (every element in the population has an equal
chance of being included in the sample)
Systematic sample: Selection from the population at a
uniform interval of a listed order.
Stratified sample: Classifying the elements in the target
population into separate groups on the basis of
characteristics associated with the variables being
studied
The objective is that the several strata be different from
one another with respect to the variables being studied, but that
the element within a given stratum be similar to one
another
 Probability Sampling Techniques……
Cluster sample: A sample in which the elements already
occur in groups, and it is the groups (cluster) that are
randomly sampled. (Often used when elements in the
population are not easily identified individually but are
grouped together (clustered) and more easily identified
as members of the cluster)

[If all elements in the selected clusters are included in the

sample, the procedure is called Single – stage sampling.
Multi – stage sampling involves further sampling
processes within each stage: eg. region → district →
community]
 Probability Sampling Techniques……

[Cluster sampling – diversity of elements within each but

the several clusters are similar to one another. Stratified
sampling – elements in each group (stratum) are similar
to one another, but the strata are structured so that they
are different].
Exit poll: Probabilistic and formal. Does not require any
prior listing of the members of the universe. Units
belonging to a specific universe arrange themselves in
single files – vehicles passing a given point on a narrow
road or people coming through a door. Choose every i th
unit which goes by.
 Probability Sampling Techniques…..

Students Reading Assignments:

1 Read on the advantages/disadvantages of

choosing one over the other
2 Read on differences/similarities between
Cluster and Stratified Sampling Techniques
3 Read on the difference between Sampling and
Census
 Non-Probability Sampling
Generalizations to whole population cannot be made because it
lacks the scientific sampling approach
 Convenience samples: The researcher is doing a case study, so
he selects one or more cases that provide the best opportunity to
observe the phenomenon or relationship of interest.
 Judgment samples: Researcher uses experience as the basis for
choosing the items to be included in the sample, with the objective
being usually to make the sample as representative of the
population as possible. e.g.:
 An accountant who chooses only certain records for a sample audit
based on the judgment that these records are representative of the
records in general
 A world bank team that selects one good project they have funded for in-
depth study
 A researcher who selects a typical farming community for in-depth study
 Non-Probability Sampling…..
The cases may be selected because:
 Judged most typical
 Represent extreme ranges
 Represent a known part of the population
 Simulate or act as a proxy
 Quota Sampling
Firstly, judgment / convenience; then secondly, quota based on
grouping! Say population classified into a number of groups based on
say age. Selection now based on age grouping by quota/proportion
 Snowball Sampling
Multi-stage process! Firstly, identify certain respondents with
“characteristic” based on “information”. Later, additional respondents
selected based on referrals.
 END LECTURE
NEXT:
Choosing a Sample Size
Confidence Intervals
Estimating a Sample Size
 Choosing a Sample Size
How large should the sample be?
Obviously:

The larger the sample the smaller the sampling

error
Sampling error depends on the actual size of the
sample
Sampling error depends not only on the sample
size but also on the sample design
 Choosing a Sample Size
What are required in choosing a sample size?
 When sample data is collected and the sample mean ẋ
is calculated, that sample mean is typically different from
the population mean μ.
 This difference between the sample and population
means can be thought of as an error (sampling error).
 We compute confidence interval based on a simple
random sample of size as:
ẋ + z*σ/ √n
confidence level for an interval determines the probability that the
confidence interval produced will contain the true parameter value
 Estimating a Sample Size…..
For example: suppose a student calculated the
sample mean of maize output of farmers to be
101.82 kg/acre, with standard deviation (σ) of
0.49. The critical value (z) for a 95% confidence
interval is 1.96.
A 95% confidence interval for the unknown mean
is:
((101.82 - (1.96*0.49)) = 101.82 - 0.96 = 100.86
(101.82 + (1.96*0.49)) = 101.82 + 0.96 =102.78

So the confidence level for mean output lies in (100.86, 102.78).

 Estimating a Sample Size…..
Margin of error
E = z*σ/ √n
The margin of error is the maximum difference
between the observed sample mean and the
true value of the population mean
Rearranging this formula, we can solve for the sample
size (n) necessary to produce results accurate to a
specified confidence and margin of error:
n = [zσ/E]2
n = z2σ2/E2
 Estimating a Sample Size…..
Using the formula for our earlier example

n = z2σ2/E2

n = (1.96)2 (0.49)2/(0.05)2
= (3.8416) (0.2401) / 0.0025
= 0.9224 / 0.0025
= 368.9 approx 370
So we will require a sample size of 370 farm
outputs to generate that level of confidence
 Sample size computation…..
In farm based sampling, usually proportions of certain attributes are know (%
adopters, for example); and for a survey design based on a simple random
sample, the sample size required can be calculated according to the formula
n = z² x p(1-p) / m²

n = required sample size

z= confidence level at 95% (standard value of 1.96)
p = estimated prevalence of farm attribute in the project area
m = margin of error at 5% (standard value of 0.05)

n= 1.96² x .3(1-.3) / .05²

n = 3.8416 x .21 / .0025
n = .8068 / .0025 = 322.72 ~ 323

Most often estimated prevalence of attribute not known. Recommended

to use maximum variance of p = 0.5
 Sample size computation…..
Design Effect: In other sampling designs and studies, a design effect is
introduced—(STUDENTS TO READ ON THIS)

For example, in anthropometric survey s designed as a cluster sample (a

representative selection of villages), not a simple random sample we need to
correct for the difference in design, and the sample size is multiplied by the
design effect (D).
The design effect is generally assumed to be 2 for nutrition surveys using
cluster-sampling methodology: n x D = 323 x 2 = 646
Then allow Contingency
The sample is further increased by 5% to account for contingencies such as
non-response or recording error: n + 5% = 646 x 1.05 = 678.3 ˜ 678

CAUTION: There are several sample size determination techniques/formulae. Please

consult a Statistician for your work/study
 ANALYZING DATA
Statistical Moments
Mean, Variance
Distributions

Econometric [Regression analyses]

OLS / 2SLS / 3SLS
Maximum Likelihood Estimation Methods
Quantal Choice Methods