Answer:-
There are various types of sampling plans which are usually divided into based on probability/random samples/where the probability of the selection of each respondent is known/ and on non-probability/non-random samples(where it is not known). In probability sampling, statistical inferences about the population can be made from the responses of the sample. For this reason, probability sampling is sometimes referred to as representative sampling. The sample is taken as representative of the population. In non-probability samples, you cannot make such statistical inferences. It may still be possible to say something sensible about the population from non-probability samples-but not on the same kind of statistical grounds.
Besides this broad classification, a number of methods are used for drawing samples, and they can be grouped into the following: (1) Simple random sampling; (2) purposive sampling; (3) stratified sampling: (4)quota sampling; (5)multistage sampling;(6)convenience sampling; and (7) self selecting sampling. These methods are categorized into the two broad classifications of sampling techniques. Here, discussion of each method will be made classifying them under the umbrella of the broad classification. Furthermore, other types of probability and non-probability sampling methods will also be discussed.
Probability/Random sampling techniques:
As to the size of a sample, while probability samples allow you to generalize from sample to population, such generalizations are themselves probabilistic. The larger the sample, the lower the likely error in generalizing may be. Probability samples are classified into the following five types of sampling methods:
a. Simple random sampling-This involves selection at random from the sampling frame of the required number of persons for the sample. If properly conducted, this gives each person an equal chance of being included in the sample, and also makes all possible combination of persons for a particular sample size equally likely. So, random sampling is the form applied when the method of selection assures each element or individual in the universe an equal chance of being chosen. It is more suitable in more homogeneous and comparatively larger groups. A random sample can be drawn either by lottery method or by using Tipett’s number or by grid system or by selecting from sequential list.
b. Systematic sampling- This involves choosing a starting point in the sampling frame at random, and then choosing every nth person. Thus if a sample of fifty is required from a population of 2,000, then every fortieth person is chosen. The problem of simple random and systematic samplings is, that both require a full list of the population, and getting this list is often difficult.
c. Stratified random sampling- This involves dividing the universe or population into a number of groups or strata, where members of a group share a particular characteristic or characteristics(e.g. stratum A may be females; stratum B males). There is then random sampling within the strata. It is usual to have proportionate sampling. It may sometimes be helpful to have dis-proportionate sampling, where there is an unequal weighting. It is possible to combine stratification with systematic sampling procedures. It is the combination of both random sampling and purposive selection. In the selection of strata, we use purposive selection method, but in selecting actual units from each stratum, random method is used.
d. Cluster/Area sampling- This involves dividing the population into a number of units, or clusters, each of which contains individuals having a range of characteristics. The clusters themselves are chosen on a random basis. The subpopulation within the cluster is then chosen. This tactic is particularly useful when a population is widely dispersed and large, requiring a great deal of effort and travel to get the survey information.
An example might involve school children, where there is initially random sampling of a number of schools, and then testing of all the pupils in each school. This method has the valuable feature that it can be used when the sampling frame is not known (e.g. when we do not have full list of children in the population, in the above example).
e. Multistage sampling- This is an extension of cluster sampling. This method is generally used in selecting a sample from a very large area. It involves selecting the sample in stages, i.e. taking samples from samples. Thus one might take a random sample of schools, then a random sample of the classes within each of the schools, and then from with in selected classes choose a sample of children. As with cluster sampling, this provides a means of generating a geographically concentrated sampling.
It is also possible to incorporate stratification into both cluster and multistage sampling. Judging the relative efficiencies of these more complicated forms of sampling, and their relationship to the efficiency of simple random sampling, is difficult, and if you are expending considerable resources on a survey it is worth seeking expert advice.
Non-probability/Non-random Sampling techniques:
In probability sampling it is possible to specify the probability that any person (or other unit on which the survey is based) will be included in the sample. Any sampling plan where it is not possible to do this is called ‘non-probability sampling`. Small-scale surveys commonly employ non-probability samples. They are usually less complicated to set up and are acceptable when there is no intention or need to make a statistical generalization to any population beyond the sample surveyed. They typically involve the researcher using his judgment to achieve a particular purpose, and for this reason are sometimes referred to as purposive samples.
A wide range of approaches has been used. The first two, quota and dimensional sampling, basically try to do the same job as a probability sample, in the sense of aspiring to carry out a sample survey which is statistically representative. They tend to be used in situations where carrying out a probability sample would not be feasible , where, for example, there is no sampling frame, or the resources required are not available. Their accuracy relies greatly on the skill and experience of those involved. The types of non probability sampling methods will be presented in short as follows:
a. Quota sampling- Here the strategy is to obtain representative of the various elements of a population, usually in the relative proportions in which they occur in the population. Quota sampling is a special form of stratified sampling. According to this method, the universe is first divided into different strata. Then the number to be selected from each stratum is decided. This number is known as quota.
b. Dimensional Sampling– It is an extension of quota sampling. The various dimensions thought to be of importance in a survey are incorporated into the sampling procedure in such a way that at least one representative of every possible combination of these factors or dimension is included.
c. Convenience sampling-It involves choosing the nearest and almost convenient persons to act as respondents. The process continues until the required sample size is reached. It is sometimes used as a cheap and dirty way of doing a sample survey. You do not know whether or not findings are representative. This is probably one of the most widely used and least satisfactory methods of sampling.
This method is generally known as unsystematic, careless, accidental or opportunistic sampling. According to this system, a sample is selected according to convenience of the field workers or researchers. The convenience may be in respect of availability of source list and accessibility of the units. It is used when universe or population is not clearly defined, sampling unit is not clear or a complete source list is not available.
d. Purposive sampling- The principle of selection in purposive sampling is the researcher’s judgment as to typicality or interest. A sample is built up which enables the researcher to satisfy his/her specific needs in a research project. Accordingly, when the researcher deliberately or purposively selects certain units for study from the population it is known as purposive selection. In this type of selection the choice of the selector is supreme and nothing is left to chance. It is more useful especially when some of the units are very important and, in the opinion of the researcher, must be included in the sample.
Factors to be considered while drawing sample
In selecting ways of choosing samples for the collection of social and economic data, the best method for any inquiry will depend on both the nature of the population to be sampled, the time and money available for investigation, and the degree of accuracy required. It should, however, be emphasized that a sample ought to be representative of the population under study. Essentially, inference from sample to populations is a matter of confidence that can be placed in the representativeness of the sample. A sample is representative to the degree to which it reflects the characteristics of population.
It must also be stressed that the representativeness of a sample is difficult, if not impossible, to check. It depends upon the degree of precision with which the population is specified, the adequacy of the sample and the heterogeneity of the population. Confidence in the representativeness of a sample is increased if the population is well defined. In another way adequacy of the sample is also an important consideration in case a very small sample is taken. To be adequate a sample must be of sufficient size to allow the researcher to have confidence in the inference. Finally, it must also be stated that representativeness depends on the degree of homogeneity of the population. The more alike the units of the population, the smaller the sample can be and still be representative. To choose a representative sample is the most difficult exercise in the sampling process. The majority of persons are subject to conscious or unconscious bias or prejudice which causes them to choose a sample which is unrepresentative in some respect.
The most popular and commonly used is the simple random sampling. The other more complex methods include stratified random sampling, proportionate stratified random sampling, disproportionate stratified random sampling, and area or cluster sampling.
To conclude, it is clear that though the procedure of selecting a sample differs according to the type of the sample selected, certain fundamental rules remain the same. These include:
(1) The universe or population must be defined precisely;
(2) Before drawing a sample, the unit of the sample should be defined;
(3) the appropriate source list which contains the names of the units of universe or population from which the sample is to be selected should be prepared before hand in case it does not already exist, and
(4) The size of the sample to be selected should be pre-determined.
—————————–
*Team, www.deepakmiglani.com