SAMPLING AN INTRODUCTION
 |
| Sampling Method in Research: Random Samples |
Researchers
often try to make inferences about the population on the basis of results from
a survey sample. To draw samples from populations, researchers must first
decide the population. Suppose that you were interested in studying the degree
to which housewives in India rely on television commercials in preference to
traditional methods. All the housewives in the country would constitutes the
population.
Take
another example. You are interested to study the degree to which the news items
in national dailies are slanted on a particular topic or party. In this case
all the news items of all the national daily newspapers will be the population.
Researchers define population as having certain characteristics.
Sampling
is the process of selecting units from a population. The total punctuation
under study is called the ’universe' of the study. This practice is required as
an alternative to census where you have to survey the entire population i.e.
data is collected from each unit. This is chosen in cases where the size of the
population is very large.
However,
the conclusion of the study can be good only when the collection of data is
done through logical reasoning behind choices. So we need to understand and
study closely the population under study and how samples are drawn from that
population in order to study the whole characteristics of that population.
A
population is the universe of events from which the sample is drawn. In other
words it is all the units about which the information is sought. The
researcher, on the basis of certain characteristics, defines populations.
Though population may be quite broad as all people or all news items in all
newspapers, they may be defined quite narrowly, such as television commercials
that appear during network prime time, i.e. between 7p.m -11p.m
Sample
is a part of the population selected for a particular research study.
Researchers rarely sample all the events or units, but rely on a portion of all
data to draw conclusions. Sampling may have the following dimension in a study
of communication research:
- In sampling events are selected from the population to be included in the study.
- The results of the study are interpreted to test hypothesis and in order to estimate parameters of the population from sample data.
The
individual elements or events in the population are assigned numbers. These
numbers from population are called units. A sample is simply a selection of
units from the population. The collection of selected individuals or events is
called the sample. A statistic is a number computed from a sample. The
statistics that reflect the features of the sample are called sample
characteristics Communication researchers gather statistics from the sample to
determine the properties of the population.
BIAS IN SAMPLING:
The
process of selecting sample units from the population has to be objective and
without bias. While sampling there is a tendency for the researcher to err.
This is referred to as bias in sampling. Certainly, researchers work to
eliminate and minimize bias. Opinion poll predictions and television rating services
are supposed to be accurate, for this elimination of bias in sampling is a
precondition. Controlling bias is critical as accuracy does not occur by accident.
Representative
sample: The goal of effective sampling should be that a good sample must be
representative of the population and big enough to permit reasonable analysis of
data. A representative sample is one that accurately reflects characteristics
of the population from which it is drawn,
SAMPLE SIZE:
"How
big should a sample be?" This question is crucial for the research
students. Generally researchers collect the sample large enough to make
reasonable interpretations. Yet, large sample size is not enough to prove that
a sample is representative of the population. The students of communication
research should try to gather a reasonably sized sample in order to make a good
and representative study.
The
size of the sample should be determined keeping in view the following factors:
- Degree of accuracy required
- Time available for completion of the study
- Manpower available
- Finances available
- Subject matter of research
There
cannot be an ideal proportion between the size of the size of the universe and
the size of the sample. ln most cases sample size is governed by the
above-mentioned factors. However, it may also be noted that very small sample
may give distorted results. At the same time, very large sample may also be wastage
of resources.
SAMPLING ERROR:
AS
we study a definite sized sample and not the entire population some error is
bound to occur in telling the characteristics of the population exactly.
Sampling
error may be defined as the degree to which sample attributes differ from
population characteristics on certain measures. If we study the entire
population then there is no sampling error at all. But we also know that it is
impossible to study the entire population over a fixed period of time and we are
bound to resort to sampling. Larger the size of the sample lesser will be the
sampling error.
Margin
of error is the amount of sampling error associated with the sample. If we
collect a big sample this margin of error can be reduced and our research
reports may be very accurate. In good sample studies, sample error is generally
indicated. For example, if in a population 40 percent of the households own
television sets, but a sample study may reveal that 39 percent households have
television sets in their homes. This is sampling error, but from research point
of view 39 percent and 40 percent is not a big difference. In research study it
may be indicated as + 1%.
FORMS OF SAMPLING
Broadly,
sampling can be done in two ways:
- Probability sampling (Random Sampling) and
- Non-probability sampling (Non-Random Sampling)
Probability
sampling is more commonly known as Random Sampling. Non-probability sampling is
called non-random sampling.
Probability Sampling (RANDOM SAMPLING):
Random sampling plays an important part in research. In this form of sampling
the selection of sample is done in such a way that each event in the population
gets equal chance of selection. Random sampling is taken for ail statistical
tools, which are applicable to data.
The
distributions of randomly occurring events can be used to figure the odds that
a sample truly reflects the characteristics of the population. Random sampling
is not haphazard, unsystematic or accidental. However, in research random means
every unit gets equal chance of selection.
Random
sampling is considered as a systematic and most scientific means of studying
the population. Random sampling consumes a lot of time and most researchers
want shortcuts. But the shortcuts throw off the whole sample and leads to
faulty results.
METHODS OF RANDOM SAMPLING:
- Simple Random sampling
- Stratified random sampling
- Cluster sampling
- Systematic sampling
Simple Random Sampling: In this type of
random sampling the selection of data is done in such a way that each event
(individuals) gets an equal chance for selection. It may be done by way of
pulling names out of a container. Numbers are assigned to each individual
events and a lot can be drawn randomly or using a random number table the
numbers to be included in the final selection can be drawn. This increases the
representativeness and sampling error can be easily computed.
Stratified random sampling: Population
is divided into different strata based on the known proportions or properties
and random sampling is completed within each group in the population. As in
simple random sampling this method is also time consuming but allows analysis
by sub division of strata and the disproportionate representation of the sub
divisions of the population is also prevented.
Say
in a study of the opinion of men and women on certain issues of a particular
place. The member of men and women are identified. After deciding the number of
men and women to be taken for study a simple random sample is drawn from each
sub group of the population stratification (the population is stratified as men
& women).
For
a study on an industry strata can be divided into managers, superiors, skilled
workers, and unskilled workers. For a study on rural youth the strata can be
student youth, non-student youth, rich-, medium-, or poor youth.
But
it should be noted that in a study based on stratified sampling, results should
be drawn for each stratum separately and various strata should not be merged
for the entire population. This will give you erroneous results.
Custer Sampling: In cluster sampling
groups of events or areas (clusters) are taken as a unit (rather than taking
single individual events as units) and an actual sample is drawn from them.
This method is considered as a practical solution to the problems of gaining
access to many settings and the cost of sampling is minimized in large-scale
surveys.
However,
this sampling is disadvantaged by the requirement of larger samples and weights
for each strata or each individual event may be difficult to know in many
settings. Results cannot be taken as representative for the entire population.
Systematic Sampling: This is a commonly
used method in which cluster sampling and stratified sampling are combined.
Every n-th event or cluster in the
population is taken for study and a systematic sampling is done among the events
or clusters thus selected.
NON-SAMPLING ERROR:
We
know that some errors are bound to occur whatever method we used for sampling.
Errors due to sampling factors (Sampling error) usually get the most attention.
However, there are a lot of factors, which affect a study. These factors (or
the sources of error) are critical as there are so many of them. The
non-sampling errors may include.
- Refusal by respondents for interview
- Intentional lying by respondent
- Prevalent opinion
- Poor recall
- Subculture of respondents
- Miscommunication
- Coding errors
- Image of the interviewers
- Recording error
- Misunderstanding of questions being asked
NON-PROBABILITY SAMPLING (NON-RANDOM SAMPLING)
This
form of sampling is applicable where random sampling is not possible.
Non-random sampling may be classified as:
- Convenience sampling,
- Quota sampling
- Purposive or known group sampling, and
- Snowball sampling
Convenience sampling: ln convenience
sampling no attempt at randomization is made. Here selection of
respondents/events depends upon the availability. Although economical in nature
the computation of bias is not possible in this case and the generalisation to
the population is out of question.
Quota sampling: This method of sampling
attempts that important parts of the population are not omitted and samples are
defined based on the known proportions within the population and non-random
sampling is completed within each group.
Purposive or Known group sampling: This
is a convenient and economical sampling method when key population
characteristics are identified. Here the selection of respondents is from
groups that are known to possess particular characteristics under investigation.
However, in this case also the generalization to the population is also not
possible.
SnowballI Sampling: Snowball sampling
is highly useful in studies where population units are not well defined and
thus cannot be listed. The selection of respondents is based on referrals from
initial informants. In this case the respondent is requested to refer the
researcher to other individuals in the group. Examples of not so well defined
population are members of underworld organization, prostitutes’ criminals, AIDS
patients, users of a particular brand, etc.
However,
studies depending on snowball sampling provide broad features of the population
and cannot be considered as actual research.
PRACTICES IN RANDOM SAMPLING:
As
we know that all the forms of sampling random or non-random have advantages and
disadvantages depending upon the type, nature of the study, we require selecting
a suitable method. We also know that the study of research is riddled with choices
and compromises. This would in the best interest of the researcher that all the
pros and cons are zoomed in and adopt so that they can produce good research
results. Randomization is a pain in the neck so far as the credibility of the
study is concerned.
