Unit 6: Sampling

6.1 Introduction

 Whatever your research question(s) and objectives you will need to consider whether you
need to use sampling. Occasionally, it may be possible to collect and analyse data from every
possible case or group member; this is termed a census.

 However, for many research questions and objectives it will be impossible for you either to
collect or to analyse all the potential data available to you owing to restrictions of time,
money and often access.

 Sampling techniques enable you to reduce the amount of data you need to collect by
considering only data from a subgroup rather than all possible cases or elements.

 Some research questions will require sample data to generalise statistically about all the cases
from which your sample has been selected. For example, if you asked a sample of consumers
what they thought of a new chocolate bar and 75% said that they thought it was too
expensive, you might infer that 75% of all consumers felt that way. Other research questions
may not involve such generalisations.

6.2 Core Issues – Population, Sample, and Sampling

 A population refers to all elements of a clearly defined body of people, events or objects. A
population may be defined in terms of four attributes, namely content, units, extent and time.
Thus, in defining youth as a population nationally, we might consider all persons 15 to 24
years old (content), in households (units) in the country of Belize (extent) in 2004 (time).

 The above approach to population definition will yield a finite population, that is one in
which it is possible to count all the elements or units. By contrast, in an infinite population,
there is no limit to the number of elements it may contain. For example, all Belizean women
in the country’s labour force over an unspecified period of time is an infinite population.

Page 1 of 8
 It is far more preferable that researchers work with finite as opposed to infinite populations.

 The three types of populations are:

o Target population:- This is another name for the finite population that the researcher
wishes to study.
o Frame population:- This is also known as the sampling frame. This refers to a list
that the researcher either develops or acquires and which, theoretically, contains all
the units of the target population. Typical sampling frames include telephone
directories, electoral lists, etc. The problem with most lists that are used as sampling
frames is that they are hardly ever complete or up-to-date. Assume that a researcher
uses a telephone directory as the sampling frame for conducting a national survey,
what problems to expect? As far as possible, researchers must aim to use a sampling
frame that has the closest possible match with the target population. A mismatch
between the frame population and the target population can be a major source of error.
o Survey population:- This comprises all the selected sampling elements (units)
included in the study. An issue that can arise in relation to the survey population is
that of non-response. Selected sample elements that have been included in the survey
population but do not yield any data are non-respondents. Non-response can be a
source of bias. While increasing sample size can compensate for the loss of numbers
through non-response, it cannot eliminate bias, since those who did not respond may
differ in characteristics from those who did.

 A distinction needs to be made between non-coverage and non-response. The former (non-
coverage) refers to a situation where potential sample elements are found to be missing from
the frame population. The latter (non-response) arises when sample elements that have been
selected for inclusion in the study do not yield data as expected.

 A sample is a subset selected from the population, and sampling is the process by which this
subset is chosen.

 Reasons for enumerating a sample:

o It may be extremely difficult or would be impracticable for you to survey the entire
population. A sample can be enumerated more quickly than a census.
o Your budget constraints prevent you from surveying the entire population. (To save
money)
o Your time constraints prevent you from surveying the entire population. (To save
time)
o It allows researchers to be able to elaborate and delve more deeply into individual
cases. This is more possible with the limited numbers covered by a sample.

 Reasons for enumerating a population:

o The population size is relatively small. In total population sampling, researchers
choose to study the entire population because the size of the population that has the
particular set of characteristics that we are interest in is typically very small.
Therefore, if you failed to include a small number of units (e.g., people) in your

Page 2 of 8
 research, a significant piece of the puzzle that you are trying to understand may be
missing.
o The population shares an uncommon characteristic(s). The characteristic shared by
the population is considered to be uncommon because this tends to explain why the
population that can be studied is very small. For example, if you were performing
case study research in a single firm of 400 employees, examining the effect of senior
manager mentorship on employee motivation, there may only be 5-10 senior
managers. In this example, the uncommon characteristic is the fact that the people
(i.e., units) of interest are all senior managers. Since the total number of senior
managers is very small, it makes sense to include all of them in your research.
o Finally, the counting and data gathering exercise that is carried out in most countries
every ten years is, by law, based on the enumeration of the entire population. This
exercise is known as a national census.

 If a sample is randomly drawn from any given population, that sample can be used for the
purpose of drawing conclusions or making inferences about the population of interest.

 Random refers to each element or unit of that population having an equal, non-zero chance of
being chosen in the sample.

 A sample is a small-scale representation of a population. Because it is smaller, it cannot be

exactly the same as the population in all aspects. Nonetheless, it (the sample) can resemble it
(the population) closely. Through sampling, researchers seek to ensure that the resemblance
is as close as possible. How close the resemblance is depends on several factors, but in
particular two, namely the size of the sample and the way in which it was selected.

 In terms of sample selection, the most effective means of ensuring this close resemblance is
to select on the basis of chance – that is, random or probability sampling.

 Finally, regardless of how close the resemblance, there is always some element of mismatch,
a discrepancy, since the sample is not an exact replica of the population. This discrepancy is
regarded as the sampling error.

 Sampling techniques can be divided into two categories:

o Probability (representative) sampling
o Non-probability sampling

6.3 Probability (Representative) Sampling Techniques

 Advantages of using probability sampling techniques:

o They are more likely to generate a representative sample.
o Researchers have greater confidence about their efforts to eliminate bias.
o Calculation of the sampling error is possible.
o Since sampling error can be calculated, it becomes possible to determine the limits of
generalizability from the sample to a given population.

Page 3 of 8
 To generate sample size, visit: http://www.raosoft.com/samplesize.html
o Margin of error:- the accuracy you require for any estimates made from your sample;
amount of error that you can tolerate.
o Confidence level:- the level of certainty that the characteristics of the data collected
will represent the characteristics of the total population; amount of uncertainty you
can tolerate. Researchers normally work to a 95 per cent level of certainty. This
means that if your sample was selected 100 times, at least 95 of these samples would
be certain to represent the characteristics of the population.

 The four types of probability sampling techniques include:

o Simple random sampling
o Systematic random sampling
o Stratified random sampling
o Cluster sampling or multi-stage cluster sampling

 Simple random sampling.

o It is the simplest and also the most basic of the probability sampling techniques.
o It ensures that every member of the population has an equal chance of being chosen.
o Any variation in the population is captured in the sample.
o It can be executed by using any of several appropriate methods, including merely
pulling numbers out of a hat.
o Sampling is done without replacement (when same number is read off a second time
it must be disregarded).
o Simple random sampling is most appropriate when the entire population from which
the sample is taken is homogeneous.
o Advantage: very simple and basic
o Disadvantage: hardly ever used because in practice, it is a very laborious exercise if
done manually.

 Systematic random sampling.

o The first step is to number all the elements in the sampling frame sequentially without
skipping.
o Calculate the sample interval, which is the standard distance between elements
selected in the sample. This interval is calculated by dividing the total number of
elements on the frame by the required sample size.
Sampling Interval = Population Size / Sample Size
o For example, if you are drawing a sample of 1,000 from a population of 10,000, using
the formula above, the sample interval is every 10th element after the random starting
point anywhere between 1 and 10 (the number between 1 and 10 can be randomly
selected). If based on random selection, element #4 becomes the starting point, the
first five units of the sample would be 4, 14, 24, 34, 44, … the last number is 9994.

 Stratified random sampling.

o Stratified random sampling is a modification of random sampling in which you divide
the population into two or more relevant and significant strata (divided into a number
of subsets) based on one or a number of attributes.

Page 4 of 8
 o This technique is applicable when it is necessary to acknowledge subgroups or strata
on the population based on one or more variables in the research question. For the
purposes of the study, it is important that each subgroup be accurately represented.
o Researchers generally use this technique when there are recognizable subgroups in
the population whose perspectives and/or attitudes ought to be reflected in the sample.
o You begin by dividing the population into strata or subgroups, then you select a
sample from each strata/subgroup using either SRS or Systematic Sampling. In
dividing the population into subgroups, seek homogeneity within the subgroup and
heterogeneity between subgroups.
o Stratified sampling techniques are generally used when the population is
heterogeneous, or dissimilar, where certain homogeneous, or similar, sub-populations
can be isolated (strata).
o As an example, consider a study that is designed to examine the eating habits of
diabetics who attend clinic at the general hospital. The researchers find that 36 of the
600 persons registered (6% of the population) are Rastafarians. For this subgroup to
be adequately represented in the sample drawn, they should constitute 6% of the
sample. Thus, in a sample of 180 (30% of the population), there should be about 10
Rastafarians.
Number of elements in the sample from subgroup = (Number of elements in
subgroup / population) x Sample size
o Based on the above example, if either SRS or systematic sampling was used, there
would be a risk of selecting no Rasta, selecting too few Rastas, or selecting too many
Rastas. Thus in the above example, the stratified sampling technique is best.
o Stratified sampling is therefore considered more superior to either SRS or systematic
sampling since it ensures the proper representation of the stratification variables in the
sample. Stratified sampling is likely to be more representative on several variables
than a SRS.
o In social research, race/ethnicity, income level, gender, age, and level of educational
attainment are some of the main variables used in stratifying a population.

 Cluster sampling or multi-stage cluster sampling.

o The previous sampling methods have dealt with procedures for sampling from lists of
elements. Such a situation is ideal. However more often than not, social research
requires selecting samples from populations that cannot be easily listed for sampling
purposes. E.g. the population of a district, all church members in Belize, all students
in Belize etc.
o In such cases, the sampling design must be much more complex. Such a design
typically involves a multistage approach, with an initial sampling of groups of
elements – clusters – followed by the selection of elements within each of the selected
clusters. This technique allows researchers to conduct studies that would otherwise be
impossible.
o Cluster sampling is a relatively complex sampling technique that may be used when it
is impossible or impractical to compile an exhaustive list of the elements or units
comprising the target population. Often, however, the population elements are already
grouped into sub-populations, and a list of those subpopulations either exists or can
be created comparatively easier.

Page 5 of 8
o Multi-stage cluster sampling then, involves repeating two basic steps: (1) listing and
(2) sampling.
 A list of primary sampling units is compiled and, perhaps, stratified for
sampling. A sample of these units is then selected.
 The selected primary sampling units are then listed and perhaps stratified. The
list of secondary sampling units is then sampled, and so forth.
o In cluster sampling, instead of selecting all the subjects from the entire population
right off, the researcher takes several steps in gathering his sample population.
o The most common cluster used in research is a geographical cluster.
o For example, a researcher wants to survey academic performance of high school
students in Belize.
 He can divide the entire population (population of Belize) into different
clusters (districts).
 Then the researcher selects a number of clusters depending on his research
through simple or systematic random sampling.
 Then, from the selected clusters (randomly selected districts) the researcher
can either include all the high school students as subjects or he can select a
number of subjects from each cluster through simple or systematic random
sampling.
 The important thing to remember about this sampling technique is to give all
the clusters equal chances of being selected.
o Types of cluster:
 One-stage cluster:- Recall the example given above; one-stage cluster sample
occurs when the researcher includes all the high school students from all the
randomly selected clusters as sample.
 Two-stage cluster:- From the same example above, two-stage cluster sample is
obtained when the researcher only selects a number of students from each
cluster by using simple or systematic random sampling.
o Advantages:
 This sampling technique is cheap, quick and easy. Instead of sampling an
entire country when using simple random sampling, the researcher can
allocate his limited resources to the few randomly selected clusters or areas
when using cluster samples.
 Related to the first advantage, the researcher can also increase his sample size
with this technique. Considering that the researcher will only have to take the
sample from a number of areas or clusters, he can then select more subjects
since they are more accessible.
o Disadvantages:
 From all the different type of probability sampling, this technique is the least
representative of the population. The tendency of individuals within a cluster
is to have similar characteristics and with a cluster sample, there is a chance
that the researcher can have an overrepresented or underrepresented cluster
which can skew the results of the study.
 This is also a probability sampling technique with a possibility of high
sampling error. This is brought by the limited clusters included in the sample
leaving off a significant proportion of the population unsampled.

Page 6 of 8
 Difference between stratified random sampling and cluster sampling:
o The main difference between cluster sampling and stratified sampling lies with the
inclusion of the cluster or strata.
o In stratified random sampling, all the strata of the population is sampled while in
cluster sampling, the researcher only randomly selects a number of clusters from the
collection of clusters of the entire population. Therefore, only a number of clusters are
sampled, all the other clusters are left unrepresented.

6.4 Non-probability Sampling Techniques

 The techniques for selecting samples discussed earlier have all been based on the assumption
that your sample will be chosen at random from a sampling frame. However, within business
research, such as market surveys and case study research, this may either not be possible (as
you do not have a sampling frame) or not be appropriate to answering your research question.
This means your sample must be selected some other way. Non-probability sampling (or non-
random sampling) provides a range of alternative techniques to select samples, the majority
of which include an element of subjective judgement.

 These techniques are those in which the probability of selecting an element to be part of the
sample is not known. These techniques may be appropriate for exploratory, observational or
even qualitative research, although the limitations of this type of sampling should be
recognized.

 For all non-probability sampling techniques, other than for quota samples (which we discuss
later), the issue of sample size is ambiguous and, unlike probability sampling, there are no
rules. Your sample size is dependent on your research question(s) and objectives.

 The three main types of non-probability sampling techniques include:

o Quota sampling
o Snowball sampling
o Convenience (haphazard) sampling

 Quota sampling. This is used intensively in commercial research. The aim is to produce a
sample that reflects a population in terms of the relative proportions of people in different
categories, such as gender, ethnicity, socio-economic groups etc. Example: If 5 males
surveyed, then there will be 5 females be surveyed too. Decisions on sample size are
governed by the need to have sufficient responses in each quota to enable subsequent
statistical analyses to be undertaken. This often necessitates a sample size of between 2000
and 5000.

 Snowball sampling. This procedure is appropriate when the members of a special population
are difficult to locate. Some examples might be homeless individuals, migrant workers or
undocumented immigrants.
o The researcher collects data on the few members of the target population that he can
locate, then ask those individuals to provide information needed to locate other
members of that population whom they happen to know.

Page 7 of 8
 o “Snowball” refers to the process of accumulation as each located respondent, and
then suggests other respondents. Because this procedure may result in samples with
questionable representation, it is used primarily for exploratory purposes.

 Convenience (haphazard) sampling. Respondents are simply available to the researcher by

virtue of their accessibility. This is the technique used by news reporters who solicit the
views of members of the public on a news item of topical interest. The main weakness of
using this technique for research purposes is that there is no way of assessing to what extent
the person interviewed, who was selected by “accident,” represents the population that they
are assumed to be a part of. Example: A reporter randomly chooses a person walking on the
street and interviews him.

 Information gained from any of the above non-probability sampling techniques should be
treated with extreme caution. Such information should not be extrapolated to generalizations
about entire populations since samples were not selected on a systematic basis.

 Other disadvantages:
o No controls for personal bias
o The ability to capture variation that exists in the wider population is not possible.
o These techniques do not allow for calculating sampling error and estimating precision.

Page 8 of 8