CHAPTER 6

SAMPLE DESIGN
 Some fundamental definitions
 Population: Is the theoretically specified aggregation of
survey elements from which the survey sample is actually
selected.
 Sampling Frame: Is the list of elements from which the
sample is drawn
 Sample: A subset or some part of a larger population
 Sample design: Is a definite plan for obtaining a sample
frame.
 Sampling: Is the process of using a small number or part of
a larger population to make conclusion about the whole
population.
 Element: Is unit from which information is collected and
which provides the basis of analysis.
 Statistic: Is a characteristic of a sample. E.g mean of sample
 Parameter: Is a characteristic of a population. E.g. mean of
population
 Sampling
Design
Census

all people or elements in a group of interest
(a population) included in a study
 Limitation of census
 very time consuming
 expensive
 minor research would not justify a census
 Advantages of census
 Reliability
 Detailed information
Sampling theory
 Is the study of the relationship existing
between a population and sample drawn from
the population.

 It is concerned with estimating the property of

the population from those of the samples and
also with gauging the precision of the
estimate.
Sampling theory Objectives
 Sampling theory is design to attain one or more of the
following objectives
 Statistical estimation: Sampling theory helps in

estimating unknown population parameters

from knowledge of statistical measurement on
sample studies.
 Testing of hypothesis: It enables us to decide

whether to not accept or to reject the stated

hypothesis.
 Statistical inference: Sampling theory helps in

making generalization about the population from

the studies based on samples drawn from it.
Advantages of Sampling
 Complete enumeration of all sample units in the
entire universe is often unnecessary to obtain
reasonably accurate results.

 An examination of the entire population is often too

costly, too time-consuming, and impractical (if not
impossible).

 In the case of destructive testing, the sample

elements or units must be destroyed or must be
consumed to obtain necessary measurements.
 Sampling technique is used under the
following conditions
 Vast data: When the number of units is very
large, sampling technique must be used.
 Infinite population
 When census is impossible
 When homogeneity is high
 CHARACTERISTICS OF A GOOD SAMPLE
DESIGN
(a) must result in a truly representative sample.
(b) must be such which results in a small sampling error.
(c)must be viable in the context of funds available for the
research study.
(d)must be such so that systematic bias can be controlled
in a better way.
(e)should be such that the results of the sample study can
be applied, for the universe with a reasonable level of
confidence.
Steps involved in sample planning

 Defining population

 Census Vs Sample

 Sampling Design

 Sample Size

 Execute Sampling Process

Defining the population
 This implies specifying the subject of the study that
involves identifying which elements (items) are included,
as well as where and when.
 If the research problem is not properly defined then
defining population will be difficult.
 E.g. What is your population of interest?
 To whom do you want to generalize your results?

 All doctors

 School children

 Taxi drivers

 Women aged 25-45 years

 Other
Sampling
Design
A sample:
 is a subset of selected from a
people
population - an example is an opinion poll

 we may want our to

sample representative of the be
value
population,
we calculate from the sample (statistic)
i.e. any
is correct as estimates of the population
value (parameter)
Sampling
Design
specify population Nonprobability sampling
convenience sample; quota
sample; judgement sample
select method of Probability sampling
sampling simple random sample; stratified
sample; cluster sample; multi-
stage sample; proportionate/
determine sample size disproportionate sample

method of collecting observation; interviewing;

questionnaire
data from sample

timing of data collection

data analysis
Sampling
Design …….
 3 factors that influence sample representativeness
 Sampling procedure
 Sample size
 Participation (response)

 When might you sample the entire population?

 When your population is very small
 When you have extensive resources
 When you don’t expect a very high response
SAMPLING BREAKDOWN
SAMPLING…….

STUDY POPULATION

SAMPLE

TARGET POPULATION
 Types of
Samples
Probability (Random) Samples
 Simple random sample
 Systematic random sample
 Stratified random sample
 Multistage sample
 Multiphase sample
 Cluster sample
 Non-Probability Samples
 Convenience sample
 Purposive sample
 Quota
Probability and Non probability Sampling
 Probability Sampling
 The distinguishing characteristic of probability sampling
is that one can specify that each element of the
population has a probability of being included in the
sample - we can estimate population parameters from
the sample statistics

 Non-probability sampling
 Is there is no way of specifying the probability of each
element’s inclusion in the sample - there is no assurance
that every element has some chance of being included
 often used when a population cannot be precisely
defined and when a population list is unavailable
Probability Sampling
 systematic procedures, not judgment, to select sample
 ensures every element from the population has a
known probability of being selected for the sample
 can evaluate the amount of error in data caused by
using a sample rather than census, i.e. a statistical
relationship exists between sample estimates and the
population
 not affected by researcher likes and dislikes
 more expensive and more difficult to obtain data
Types of Probability Sampling
1. Simple Random Sample:
 Items selected from population such that each has an

equal and known probability of selection

 can allocate a number to each item and sample can

be drawn using random number tables or lottery

method.
 Applicable when population is small, homogeneous
& readily available.
e.g. a bank randomly selects from a list of 800 customers a
sample of 200; each list member has a 200/800 (25%)
chance of being included in the sample
- known probability is n/N, where

- n is the sample size and

- N population size
Types of
… 2. Systematic Random Sample
•relies on arranging the target population according
to some ordering scheme and then selecting
elements at regular intervals through that ordered
list.
• a random starting point selected
• every Kth unit in the frame is selected,

•where: K = Population Size/ Sample Size

Types
of…
3. Stratified Sample
• a stratum = a population segment having one or more
common characteristics (e.g. companies with between
10 and 50 employees)
• When the population is heterogeneous overall, but within
it there are homogeneous (strata)
populations population is stratified. the
• population divided into strata
• the basis for this division is judgment
• may help to ensure a sample, thus
representative reducing sampling error
Types
of…
 One of the main reasons for using a stratified sample
is that stratifying has the effect of reducing sampling
error for a given sample size to a level lower than that
of an simple random sample of the same size.
 This is so because of a very simple principle: the more
homogeneous a population is on the variables being
studied, the smaller the sample size needed to
represent it accurately.
 Stratifying makes each sub-sample more
homogeneous by eliminating the variation on the
variable that is used for stratifying.
Probability Sampling
Proportionate / Disproportionate Sampling
 proportionate - an equal % of respondents
sampled for each strata
 disproportionate - a small important stratum is
over-sampled but restored to due weight in
overall results
Types
of…
4. Cluster Sample
 Another modified random sample design --

requires that the sample unites be grouped in

clusters in the universe.
 Not grouped by homogeneous strata in the

population.
 population divided into clusters before
sample taken
 randomly choose clusters and sample all elements within
each cluster
 low cost for a given sample size
Difference Between Strata and Clusters
 Although strata and clusters are both non-overlapping
subsets of the population, they differ in several ways.

 All strata are represented in the sample; but

only a subset of clusters are in the sample.

 With stratified sampling, the best survey results occur

when elements within strata are internally
homogeneous.

 However, with cluster sampling, the best results occur

when elements within clusters are internally
heterogeneous.
Types
of…
5. Multi-stage Sample
A probability sample taken in multiple stages if survey
population is large and widely dispersed.
 The selection procedure takes place in a hierarchy of
stages.
 First primary sample unit
 Second second sample
 Third unit tertiary sample
 . . . . . unit
 last final (or ultimate) sample unit
 Comparisons of probability sampling techniques
Sampling Cost Degree Advantage Disadvantage
of use
technique
Simple High cost Moderate Easy to analyze data and Requires sampling frame;
random sampling error Large errors for same
sample size compared to
stratified; High cost if
respondents are dispersed

Systematic Moderate Moderate Simple to draw sample Requires sampling frame

to high
cost

Stratified High cost Moderate Ensures group Requires information for

representation; stratification variable;
Comparison among Sampling frame needed
strata possible; reduce
variability

Cluster Low cost Frequent Lowers field cost; Larger errors for same
Possible to estimate Xics sample size compared to
of clusters other techniques
 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
 Comparisons of non-probability sampling techniques
S Sampling Cost Degree of Advantage Disadvantage
N technique use

1 Convenience Very low Extensively No need for list of Samples are

used population unrepresentative;
Making inference
beyond sample risky

2 Judgmental Moderate Average Sample guaranteed Bias due to expert’s

cost to meet a specific beliefs make samples
objective; useful for unrepresentative;
certain types of Making inference
forecasting beyond samples risky

3 Quota Moderate Very Requires no need Introduces bias in

cost extensively for list of population researcher’s
used classification of
subjects; Making
inference beyond
sample risky

4 Snowball Low cost Used in Useful in locating High bias because

special rare population sample units are not
situations independent; Inference
beyond samples risky
Sample Size
 sample size (number of elements in sample)
and precision of the study are directly related
 The larger the sample size the higher is
the accuracy.
 The sample size determination is
statistical activity, purely which needs
knowledge. statistical
 the adequate size of a sample is properly
estimated by deciding what level of accuracy
is expected (i.e. how large a standard error or
sampling error is acceptable)
 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.
Sample Size
There are a number of sample size determination methods
 Personal judgments: The personal judgment and

subjective decision of the researcher in some cases can

be used as a base to determine the size of the sample.
 Budgetary approach is another way to determine the

sample size. Under this approach the sample size is

determined by the available fund for the proposed study.
 E.g., if cost of surveying of one individual or unit is 30 birr

and if the total available fund for survey is say 1800 birr ,
the sample size then will be determined as,
 Sample size (n) = total budget of survey /Cost of unit
survey
Accordingly, the sample size will be 60 units (Br.1800 / Br.30 per unit
= 60 units)
Traditional inferences
 This is based on rate and
precision level.
confidence
 To estimate size usingthis
sample approach have information
we need to about:
 the estimated variance of the population,
 the magnitude of acceptable error and
 the confidence interval
Cont…
Steps in estimating sample size:
1. Estimate the standard deviation of the population
1. From previous study or pilot study
2. Decide on the magnitude of error
3. Determine the confidence interval
4. Calculate sample size as follows:
 Variance or heterogeneity of the population: refers to
the standard deviation of the population parameter.
 From previous study or pilot study

 If information about variance is not available

a researcher is expected to estimate it.
 According to the rule of the thumb standard deviation is
one-sixth of the range.
 Magnitude of acceptable error: The magnitude of
error (range of possible error) indicates how precise the
study must be.
 It is acceptable error for that study. The researcher
makes subjective judgment about the desired
magnitude of error.
 E.g., to estimate the average income of household
one may allow an error says  50
Confidence interval:
 In most case (research) 95% confidence level is used.
That is, it is assumed that 95 times out of 100 the
estimate from sample will include the population
parameter.
 Once the above concepts are understood and

determined the size of sample is quite simple to

determine. It is determined based on the following
relationship.
n = (ZS/E)2 and (for unlimited population)

n = N/ (1+N (e) 2 (for known population)

Where Z is standardization value indicating a confidence
level E represents acceptable magnitude of error  an error
factor S represents sample SD or an estimate of the
population SD n represent sample size
Tables
 It is also possible to use 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.
Source: The NEA Research Bulletin (1960), Vol 38:99
40
Sampling Error and Non Sampling Error

Sampling Error
 The sampling error refers to the extent to which

the sample values on some variable of importance

to the research differ from those of the population
from which it was drawn.
 Sampling error is the difference between the result

of a sample and the result of census.

 It is the difference between the sample estimation

and the actual value of the population.

 These are errors that are created because of the

chance only
Errors
 Sampling (internal) Error

The fact that a sample was taken, the

sample statistic is expected to deviate
from the population parameter.
 The mean of the sample might be different
from the population mean by chance
alone.
 The standard deviation of the sample
might also be different from the population
standard deviation.
 Therefore, we can expect some difference
between the sample statistics and the
population parameter.
 This difference is known as sampling error.
Example
 Suppose an individual student has scored the
following grades in 10 subjects (Consider these
subjects as population); 55, 60, 65, 90, 55, 75, 88,
45, 85, 82.
 Say, a sample of four grades 55, 65, 82, and 90 are
selected at random from this population to estimate
the average grade of this student. The mean of this
sample is 73. But the population mean is 70. The
sampling error is therefore, 73 - 70 = 3.
 However, the variation due to random fluctuation
(sampling error) decreases as the sample size
increases though it is not possible to completely
avoid sampling error.
Systematic Error (non-sampling error)
 Systematic error is also called sampling bias.
 Such error can be created from errors in the sampling

procedure, and it cannot be reduced or eliminate by

increasing the sample size.
 Such error occurs because of human mistakes and

not chance variation.

 Non-Sampling (external) Error

The practical considerations in taking a sample.

recording errors
processing errors
 Generally, non-sampling errors occur in a sample
survey as well as in census survey where as the
sampling error occurs only in a sample survey.
 Preparing the survey questionnaire and handling
the data properly can minimize non-sampling error.
 as sample size increases, non-sampling errors (in
for example, data collection, non-responses,
measurement and analysis) may increase
 Maximizing accuracy requires that total study error
be minimized.
 Total error = sampling error + Non-sampling
Error
Errors
 Bias
Most insidious to detect ....
poorly defined universe
inadequate sampling design
improperly worded questions
distorted answers
Non response rate
End of Chapter
Six