SAMPLING METHOD

When taking a sample we hope it has similar features to the underlying population, and thus
provides a balanced reflection of it. The samples should be representative of the entire
population. There are three main reasons for drawing a sample instead of taking a complete
census. They are:

i) it is usually too time consuming to perform a complete census,

ii) it is too costly to do a complete census,
iii) it is just too cumbersome and inefficient to obtain a complete count of the target
population.

Basically, sampling methods can be divided into two main categories: Nonprobability
sampling and Probability Sampling.

NONPROBABILITY (NONRANDOM) SAMPLING

Nonprobability sampling is a subjective procedure in which the probability of selection for

each population unit is unknown beforehand. There are three kinds of nonprobability
sampling, namely, the convenience sampling, the judgment sampling and the quota sampling.

1. Convenience Sampling

It is a procedure in which a researcher’s convenience forms the basis for selecting a

sample of units.

Example
A professional accountant belonging to a political party wants to examine the
attitudes of local residents towards increased income taxes. A number of interviewers
are hired to give questionnaires to people entering a supermarket and ask them to
drop it into a box that is conspicuously placed in the supermarket entrance when they
have filled out the questionnaires. Obviously, the sample is not representative since
many of them are housewives who may not know the effect of increased income taxes.
Also, many shoppers are just walking off with unfilled questionnaires.

2. Judgment Sampling

Judgment sampling is similar to convenience sampling, except that it is more refined.

In convenience sampling a researcher does not exert any effort to obtain a
representative sample. Such effort is expended in judgement sampling. Specifically,
it is a procedure in which a researcher exerts some effort in selecting a sample that he
or she feels is most appropriate for a study.

Example
Rather than choosing the records that are to be audited on some random basis, an
accountant chooses the records for a sample audit based on the judgment that these
particular types of records are likely to be representative of the records in general.
Another example is to test all repairs and maintenance expense items in excess of
$10,000.00. The use of judgmental selection is widespread even by auditors who are
strong advocates of statistical methods.
3. Quota Sampling

Quota sampling is the most refined form of nonprobability sampling and is often used
in practice, especially in studies involving personal interviewing. In quota sampling,
the selection is constrained by various pre-established quotas such as age, sex, race,
etc., which try to simulate known population characteristics.

Example
In order to examine taxpayers’ attitudes toward recent government tax policies, an
interviewer is assigned to interview 100 male, self-employed persons over 45 years of
age, 80 working females aged 25-35 living with families, 50 male with home on
mortgage, and so on. Once these quotas are established, the interviewer may select
individuals who are most readily available with a working or living area in the
required time.

Quota sampling is superior to both convenience and judgment sampling in terms of

sample representativesness. The relative flexibility of quota sampling and its ability
to provide reasonably representative samples have made it a popular technique.

From the above illustration, the advantages and disadvantages of nonprobability sampling are
as follows:

ADVANTAGES:
1. It is usually simpler and cheaper to carry out than probability sampling since it does not
require the probabilistic ways to choose a sample.
2. It is useful when the population list (sampling frame) is not available.

DISADVANTAGES:
1. The interviewer is given too much discretion in the selection process.
2. There is no probabilistic way of interpreting how representative the particular sample is of
the overall population.

PROBABILITY (RANDOM) SAMPLING

Probability sampling is an objective procedure in which the probability of selection is known

in advance for each population unit. In other words, it is a type of sampling in which every
item in a population of interest has a known, and usually equal, chance of being chosen for
inclusion in the sample. The four principal methods of probability sampling are the simple,
systematic, stratified, and cluster sampling methods.

1. SIMPLE RANDOM SAMPLING

Simple random sampling is the most basic type of probability sampling for selecting a
specified number of units form a population. It is simply a procedure in which items
are chosen individually from the entire target population on the basis of chance. Such
chance selection is similar to the random drawing of number in a lottery.

Example
Suppose we want to take a simple random sample of 10-account receivable for a
population of 90 accounts, with the accounts being numbered 01 to 90. We would
enter the table of random numbers “blindly” by literally closing our eyes and pointing
to a starting position. Then we would read the digits in groups of two in any direction

Paul Chan (July,2005) 2

 to choose the accounts for our sample. If any of the selected numbers is outside the
range of 01-90, we ignore it and take another two-digit numbers. Also, the selected
numbers can appear only once. Many textbooks use the table of random number to
illustrate how to select a random sample. However, it is not practical in the real world
since computers or even calculators are more efficient than using the random number
table.

Although simple random sampling is the most basic probability-sampling method, it

has certain drawbacks in terms of its practical usefulness.

i) It requires a complete list of population units or an exhaustive sampling

frame.
ii) Unless the population list is computerized, it would be difficult to generate a
sample if the population is relatively large.
iii) Because it guarantees an equal probability of being selected, a simple random
sample can turn out to be an extreme sample.

Since there are other probability sampling methods which can overcome some of
these difficulties, simple random sampling is not very widely used in practical
settings. However, it lays the statistical groundwork that is basic to all probability-
sampling techniques and is a foundation for more advanced probability-sampling
techniques.

2. SYSTEMATIC RANDOM SAMPLING

It is an organized procedure for selecting a sample in which the items are selected
from the population at a uniform interval of a listed order. Specifically, it involves
these steps:

i) A sampling interval k is determined as follows:

Number of units in the population

K=
Number of units desired in the sample

ii) One unit between the first and kth units in the population is randomly
chosen.

iii) Then the randomly chosen unit and every kth unit thereafter are
designated as part of the sample.

Example
Suppose a company has 100 account receivables, and these accounts are coded by
number 001-100. We want to draw a systematic sample of 10 accounts. K would then
equal to 100/10 = 10. Let the first random number within 1-10 be 3. Then, the sample
will consist of the account receivables numbered as 3, 13, 23, 33, 43,53,63,73,83,93.

Example
Suppose the housing mortgage unit in a finance corporation has approved N
mortgages over a period of 20 years. These N mortgages have been numbered serially
and the order depends on the date of mortgage approval. It has been found that there
is a generally increasing interest rate over the years. In order to study this trend, a
random sample is going to be drawn. Since the mortgages have been numbered in
order with respect to time, a suitably designed systematic random sample could

Paul Chan (July,2005) 3

 ensure that the selection of data would spread evenly over the 20 years period to
reflect the real situation. By using simple random sampling, a bias sample could
possibly be produced. For instance, a majority of the sampled data which are closed
together with respect to time could be selected. These would be poor estimates of
studying a long-term trend of interest rates over a 20 years period.

Although this type of sampling is easy and inexpensive, care must be taken that no set
of items in the population recurs at set interval. For example, if every fourth product
in an assembly line is selected for quality control inspection, and if four different
machines are sequentially producing products, all products sampled may have been
manufactured by the same machine.

3. STRATIFIED RANDOM SAMPLING

It is a probability-sampling procedure in which a stratified random sample is one

obtained by separating the population elements into nonoverlapping groups, called
strata, and then selecting a simple random sample from each stratum.

The general procedure for taking a stratified sample is as follows:

Step 1: The population of interest is divided into strata on the basis of an appropriate
Population characteristic so that units within each stratum are alike and units
across strata are different.

Step 2: Calculate the proportion of the population lying in each partition.

Step 3: Split the total sample size up into the above proportions by the following
formula:

number of population units

Number of units in the stratum
allocated to = total number of units in x total sample size
a stratum the population

Step 4: Take a separate sample (usually by simple random sampling method) from
each partition based on the sample size as defined in step 3.

Example
In the confirmation of account receivable, the auditor normally expects larger error in
accounts with larger balances than in those with small balances. Consider a
population of accounts receivable of 6000 receivables. An audit wants to select a
sample of 400 account receivable. The auditor decided to use three strata and obtain
the following breakdown.

Value of receivable Number of accounts receivable

Over $2,000 3030

$500 - $2,000 2350
less than $500 620
6000

Paul Chan (July,2005) 4

 sample size 400 1
In this case the sampling fraction is = =
Population size 6000 15

Hence from each stratum we require a sample of size 1/15 x (stratum size); for
example, from the value of receivable over $2,000 stratum we take a sample of size
3030 x 1/15 = 202. In the same way we obtain the sample sizes from each stratum:

Value of receivable Number of accounts receivable

Over $2,000 202

$500 - $2,000 157
less than $500 41

400

In the above example, because the number of units selected from each stratum is
proportional to the number in the stratum, the procedure is called proportionate
stratified random sampling. Otherwise, it is disproportionate stratified sampling, say
drawing equal number (440 / 3 = 133.33 in the above example) of observations from
strata with various sizes.

There are some reasons for using stratified sampling.

i) It may accurately reflect characteristics of the population when the stratified

sample is properly designed.
ii) Information concerning individual stratum may be desired.
iii) It is easier to collect information for either physical or administrative reasons.

On the other hand, the stratified random sampling method involves an extensive
sampling frame; thus it increases costs in terms of time and manpower.

4. CLUSTER RANDOM SAMPLING

It is a probability-sampling procedure in which the population is firstly divided into

cluster or primary groups. When all of the elements in the primary groups are
included in the sample, this is called single-stage sampling. When the sampling
process involves the selection of subgroups from within the primary sample groups,
this is called two-stage or multi-stage sampling.

Usually, we use cluster sampling if

i) he sampled population is large and spread over a large geographic area,

ii) a reasonably accurate listing of the population members (sampling frame) is
not available.

Example
Suppose we want to estimate the opinions of the heads of household in a city about
the property tax but we know that many households, for one reason or another, are
not listed on the tax roll. How could we proceed?

Paul Chan (July,2005) 5

 Since the tax roll cannot be used as a frame, we could construct a frame by
numbering on a map each of the city blocks. The list of all blocks in the city would
then provide a frame for selecting a random sample of city blocks, each representing
a cluster of households. After the random sample of clusters is selected, interviewers
are sent out to contact and interview al heads of household within each cluster (block)
that appear in the sample. The information contained in the clusters is ultimately
combined to make inferences about the population of heads of household within the
city.

From the above example, the advantages of cluster sampling are clear:

i) it is often easy to form a frame consisting of clusters,

ii) it is less costly to conduct interviews within clusters than to interview
individual elements selected at random from the population.

CONCLUSION

Although a nonrandom sample can turn out to be representative of the population, there is
difficulty in assuming beforehand that it will be unbiased, or in expressing statistically the
confidence that can be associated with inferences from such a sample. The choice between
nonprobability and probability sampling for a study depends on the study’s objectives as well
as the resources available for conducting the study. It should be bear in mind that a sampling
design can incorporate more than one of the sampling methods to fit the unique circumstances
surrounding a particular research project.

Paul Chan (July,2005) 6