BUS 802 Business Research Methods

Lecture Note April 8, 2017

Dr. Olu Akintunde, CPA. PMP. CISA

The purpose of this lecture is to provide an overview of the sampling procedures available to a
researcher. The differences between the various sampling procedures are discussed and
examples are provided to illustrate the use of these procedures. The emphasis is placed on
underlying ideas and methods rather than detailed mathematical derivations. For the reader
who is interested in pursuing a more thorough approach to the topics discussed, a list of
technical references is provided in the box below.

Why Sample
If a researcher desires to obtain information about a population through questioning or testing,
he/she has two basic options:
1. Every member of the population can be questioned or tested, a census; or
2. A sample can be conducted; that is, only selected members of the population are questioned
or tested.
Contacting, questioning, and obtaining information from a large population, such as all of the
households residing in Agege local government, is extremely expensive, difficult, and time
consuming. A properly designed probability sample, however, provides a reliable means of
inferring information about a population without examining every member or element.
Often, researchers are working under strict time constraints which make conducting a
census unwieldy. For instance, national polling firms frequently must provide information
on the public's perceptions of current events or issues. These polling firms tend to limit
their national sample sizes to approximately 1,500 respondents. When properly
conducted, a probability sample of this size provides reliable information with a very
small margin of error for the whole population of the United States, which is more than 300
million persons.
A probability sample frequently is more accurate than a census of the entire population. The
smaller sampling operation lends itself to the application of more rigorous controls, thus
ensuring better accuracy. These rigorous controls allow the researcher to reduce nonsampling
errors such as interviewer bias and mistakes, nonresponse problems, questionnaire design
flaws, and data processing and analysis errors. In part, these nonsampling errors are reduced
through pretesting which allows careful testing of the survey questionnaire and procedures.
Pretesting cannot be done when conducting a census without causing possible contamination of
some of the respondents. The detail of information that can be asked in a sample is greater than
that in a census due to the cost and time constraints under which most researchers are
operating. A relatively long and difficult questionnaire can be administered to a sample more
easily than a brief questionnaire can be administered to the entire population. However, not all
samples are accurate or the appropriate vehicle for gathering information or testing a hypothesis
about a population. The following sections of this brochure will briefly discuss the merits and
disadvantages of various sampling procedures.

Sampling Methodologies
Sampling methodologies are classified under two general categories:
1. Probability sampling and

2. Nonprobability sampling.
In the former, the researcher knows the exact possibility of selecting each member of the
population; in the latter, the chance of being included in the sample is not known. A probability
sample tends to be more difficult and costly to conduct. However, probability samples are the
only type of samples where the results can be generalized from the sample to the population. In
addition, probability samples allow the re-searcher to calculate the precision of the estimates
obtained from the sample and to specify the sampling error.
Nonprobability samples, in contrast, do not allow the study's findings to be generalized from the
sample to the population. When discussing the results of a nonprobability sample, the
researcher must limit his/her findings to the persons or elements sampled. This procedure also
does not allow the re-searcher to calculate sampling statistics that provide information about the
precision of the results. The advantage of nonprobability sampling is the ease in which it can be
administered. Nonprobability samples tend to be less complicated and less time consuming
than probability samples. If the researcher has no intention of generalizing beyond the sample,
one of the nonprobability sampling methodologies will provide the desired information.

Nonprobability Samples
The three common types of nonprobability samples are convenience sampling, quota sampling,
and judgmental sampling.
A. Convenience Sampling
As the name implies, convenience sampling involves choosing respondents at the convenience
of the researcher. Examples of convenience samples include people-in-the- street interviews—
the sampling of people to which the researcher has easy access, such as a class of students;
and studies that use people who have volunteered to be questioned as a result of an
advertisement or another type of promotion. A drawback to this methodology is the lack of
sampling accuracy. Because the probability of inclusion in the sample is unknown for each
respondent, none of the reliability or sampling precision statistics can be calculated.
Convenience samples, however, are employed by researchers because the time and cost of
collecting information can be reduced.
B. Quota Sampling
Quota sampling is often confused with stratified and cluster sampling—two probability sampling
methodologies. All of these methodologies sample a population that has been subdivided into
classes or categories. The primary differences between the methodologies is that with stratified
and cluster sampling the classes are mutually exclusive and are isolated prior to sampling.
Thus, the probability of being selected is known, and members of the population selected to be
sampled are not arbitrarily disqualified from being included in the results. In quota sampling, the
classes cannot be isolated prior to sampling and respondents are categorized into the classes
as the survey proceeds. As each class fills or reaches its quota, additional respondents that
would have fallen into these classes are rejected or excluded from the results.
An example of a quota sample would be a survey in which the researcher desires to obtain a
certain number of respondents from various income categories. Generally, researchers do not
know the incomes of the persons they are sampling until they ask about income. Therefore, the
researcher is unable to subdivide the population from which the sample is drawn into mutually
exclusive income categories prior to drawing the sample. Bias can be introduced into this type
of sample when the respondents who are rejected, because the class to which they belong has
reached its quota, differ from those who are used.
C. Judgmental Sampling
In judgmental or purposive sampling, the researcher employs his or her own "expert” judgment
about who to include in the sample frame. Prior knowledge and research skill are used in
selecting the respondents or elements to be sampled.
An example of this type of sample would be a study of potential users of a new recreational
facility that is limited to those persons who live within two miles of the new facility. Expert
judgment, based on past experience, indicates that most of the use of this type of facility comes
from persons living within two miles. However, by limiting the sample to only this group, usage
projections may not be reliable if the usage characteristics of the new facility vary from those
previously experienced. As with all nonprobability sampling methods, the degree and direction
of error introduced by the researcher cannot be measured and statistics that measure the
precision of the estimates cannot be calculated.

Probability Samples
Four basic types of methodologies are most commonly used for conducting probability samples;
these are simple random, stratified, cluster, and systematic sampling. Simple random sampling
provides the base from which the other more complex sampling methodologies are derived.
A. Simple Random Sampling
To conduct a simple random sample, the researcher must first prepare an exhaustive list
(sampling frame) of all members of the population of inter-est. From this list, the sample is
drawn so that each person or item has an equal chance of being drawn during each selection
round. Samples may be drawn with or without replacement. In practice, however, most simple
random sampling for survey re-search is done without replacement; that is, a person or item
selected for sampling is re-moved from the population for all subsequent selections. At any
draw, the process for a simple random sample without replacement must provide an equal
chance of inclusion to any member of the population not already drawn. To draw a simple
random sample without introducing researcher bias, computerized sampling programs and
random numbers tables are used to impartially select the members of the population to be
sampled.
An example of a simple random sample would be a survey of County employees. An exhaustive
list of all County employees as of a certain date could be obtained from the Department of
Human Resources. If 100 names were selected from this list using a random number table or a
computerized sampling program, then a simple random sample would be created. Such a
random sampling procedure has the advantage of reducing bias and enables the researcher to
estimate sampling errors and the precision of the estimates derived through statistical
calculations.
B. Stratified Random Sampling
Stratified random sampling involves categorizing the members of the population into mutually
exclusive and collectively exhaustive groups. An independent simple random sample is then
drawn from each group. Stratified sampling techniques can provide more precise estimates if
the population being surveyed is more heterogeneous than the categorized groups, can enable
the re-searcher to determine desired levels of sampling precision for each group, and can
provide administrative efficiency.
An example of a stratified sample would be a sample conducted to determine the average
income earned by families in the United States. To obtain more precise estimates of in-come,
the researcher may want to stratify the sample by geographic region (northeast, mid- Atlantic, et
cetera) and/or stratify the sample by urban, suburban, and rural groupings. If the differences in
income among the regions or groupings are greater than the income differences within the
regions or groupings, precision of the estimates is improved. In addition, if the research
organization has branch offices located in these regions, the administration of the survey can be
decentralized and perhaps conducted in a more cost-efficient manner.
C. Cluster Sampling
Cluster sampling is similar to stratified sampling because the population to be sampled is
subdivided into mutually exclusive groups. However, in cluster sampling the groups are de-fined
so as to maintain the het-erogeneity of the population. It is the researcher’s goal to establish
clusters that are representative of the population as a whole, although in practice this may be
difficult to achieve. After the clusters are established, a simple random sample of the clusters is
drawn and the members of the chosen clusters are sampled. If all of the elements (members) of
the clusters selected are sampled, then the sampling procedure is defined as one-stage cluster
sampling. If a random sample of the elements of each selected cluster is drawn, then the
sampling procedure is defined as two-stage cluster sampling.
Cluster sampling is frequently employed when the researcher is unable to compile a
comprehensive list of all the elements in the population of interest. A cluster sample might be
used by a researcher attempting to measure the age distribution of persons residing in Fairfax
County. It would be much more difficult for the researcher to compile a list of every person
residing in Fairfax County than to compile a list of residential addresses. In this example, each
address would represent a cluster of elements (persons) to be sampled. If the elements
contained in the clusters are as heterogeneous as the population, then estimates derived from
cluster sampling are as precise as those from simple random sampling. However, if the
heterogeneity of the clusters
is less than that of the population, the estimates will be less precise.
D. Systematic Sampling
Systematic sampling, a form of one-stage cluster sampling, is often used in place of simple
random sampling. In systematic sampling, the researcher selects every nth member after
randomly selecting the first through nth element as the starting point. For example, if the
researcher decides to sample every 20th member of the population, a 5 percent sample, the
starting point for the sample is randomly selected from the first 20 members. A systematic
sample is a type of cluster sample be-cause each of the first 20 members of the sampling frame
defines a cluster that contains 5 percent of the population.
A researcher may choose to conduct a systematic sample instead of a simple random sample
for several reasons. Systematic samples tend to be easier to draw and execute. The researcher
does not have to jump backward and forward through the sampling frame to draw the members
to be sampled. A systematic sample may spread the members selected for measurement more
evenly across the entire population than simple random sampling. Therefore, in some cases,
systematic sampling may be more representative of the population and more precise.
One of the most attractive aspects of systematic sampling is that this method can allow the re-
searcher to draw a probability sample without complete prior knowledge of the sampling frame.
For example, a survey of visitors to the County's publications desk could be conducted by
sampling every 10th visitor after randomly selecting the first through 10th visitor as the starting
point. By conducting the sample in this manner, it would not be necessary for the researcher to
obtain a comprehensive list of visitors prior to drawing the sample.
As with other types of cluster sampling, systematic sampling is as precise as simple random
sampling if the members contained in the clusters are as heterogeneous as the population. If
this assumption is not valid, then systematic sampling will be less precise than simple random
sampling. In conducting systematic sampling, it is also essential that the re-searcher does not
introduce bias into the sample by selecting an inappropriate sampling interval. For instance,
when conducting a sample of financial records, or other items that follow a calendar schedule,
the researcher would not want to select “7” as the sampling interval because the sample would
then be comprised of observations that were all on the same day of the week. Day-of-the-week
influences may cause contamination of the sample, giving the researcher biased results.

Summary
Sampling can be a powerful tool for accurately measuring opinions and characteristics of a
population. However, there is a genuine potential for misuse of this tool by researchers who do
not understand the limitations of various sampling procedures. The differences between
nonprobability and probability sampling procedures are often difficult to discern but are
extremely important for determining how the results of the research can be used. Nonprobability
sampling techniques can provide valuable information but the results cannot be generalized to a
larger population nor can statistics indicating the reliability of the results be calculated. Well
conducted probability samples provide the researcher with the ability to gather information from
a relatively small number of members of a large population and accurately generalize the results
to the entire population. In addition, probability samples enable the re-searcher to calculate
statistics that indicate the precision of the data.

Glossary
 Bias (error): Distorted or unreliable survey results. All surveys contain some bias. Bias is
increased when the respondents (persons answering the survey) are not representative of the
population being questioned, when questions are poorly written or misunderstood, and when the
researcher uses inappropriate techniques to analyze the data.

 Census: A study using all available elements (members) of a population.

 Data: The collection of observations and information resulting from the survey process.

 Element (member): The basic unit about which survey information is sought (i.e., person,
business, household, car, dog, et cetera).
 Heterogeneous: A population whose elements have dissimilar characteristics. Heterogeneity
is the state of being dissimilar.

 Homogeneous: A population whose elements have similar characteristics. Homogeneity as

the state of being similar.
 Instrument: The tool or device used for survey measurement; usually a questionnaire.

 Nonresponse: Unit nonresponse refers to the refusal of persons selected to be sampled to

participate in a survey (i.e., person does not return the mail questionnaire). Item nonresponse
refers to selected questions left unanswered by the person surveyed.
 Population: The universe or collection of all elements (persons, businesses, et cetera) being
described or measured by a sample.
 Pretest: An initial evaluation of the survey design by using a small, subsample of the intended
population for preliminary information.

 Questionnaire: A measuring device used to query a population/sample in order to obtain

information for analysis.

 Respondent: An element or member of the population selected to be sampled.

 Sample: Any portion of the population, less than the total.

 Sampling Frame: An exhaustive list of all members of the population from which a sample can
be drawn.
 Survey: A process of inquiry for the purpose of data collection and analysis using observation,
polls, questionnaires, and/or interviews.