ADVERTISEMENTS:
After reading this article you will learn about the combinations of probability and non-probability sampling.
If sampling is carried out in a series of stages, it is possible to combine probability and non-probability principles in one sampling design. One or more stages of sampling can be carried out according to the probability principle and the remaining stages according to the non-probability principle.
To take an example, the investigator may begin by selecting clusters using cluster (probability) sampling strategy but, at the final stage, he may select classes of elements as quota samples.
ADVERTISEMENTS:
Thus, the sampler may select a probability sample of districts in a state; within each of these districts, he may choose a probability sample of development blocks and finally, within each block he may select quota samples controlled for the stages of community development, i.e., I, II, III, etc.
The advantage of such a design is that the major economies of quota sampling occur in obtaining the particular cases for the sample. It is relatively less expensive to select by recourse to the probability principle, areas within which the final stage of sampling will take place.
There is some evidence to prove that quota samples taken in selected areas are more successful in controlling certain variables than is the case when the control of these variables depends on the judgements of interviewers or observers. Combining probability and non-probability procedures in certain instances, may involve an opposite strategy.
The investigator may take a probability sample of elements within a non-probability sample of areas; the areas are selected as a purposive or judgment sample. The districts (in the above example) may be selected on the ground that these have been particularly successful in reaching the developmental targets (or the reverse) and from each of these, the sampler then selects a probability sample of the developmental blocks.
ADVERTISEMENTS:
The typical districts selected purposively may be regarded as defining a population. If a probability sampling is completely applicable and hence one can estimate the degree of confidence that may be placed in the assumption that findings of the sample are a good representation of the ‘population’ returns.
The researcher can then generalize the inferences based on this restricted sub-population to the national population, subject to the assumption that the typical districts are still typical of their respective stages. So long as and to the extent that this assumption is valid.
Let us now discuss at some length the special applications of non-probability sampling. It was suggested earlier that the major advantages of non-probability sampling procedures are convenience and economy. Investigators continue to use non- probability sampling methods and justify their use on grounds of practical experience, expediency and facility.
They may, of course, in the same breath concede the theoretic superiority of the probability sampling. Many practical samplers, however, argue that in many cases, the superiority of the probability sampling is only ‘on paper’ or ‘notional.’ They point out that many a time, the way the probability sampling plans are actually implemented, the theoretical advantages of probability sampling get nearly nullified.
There can be many slips in carrying out the probability sampling plan. For instance, some of the cases selected in the sample may refuse to be interviewed or not be available, interviewers may omit some of the questions in the process of interviewing, compromises may be effected by allowing interviewers to substitute other respondents when the originally selected cases are not found at home and so on.
The sample actually interviewed, therefore, may not be a probability sample of the universe in the strict sense of the term.
Moreover, there are circumstances in which probability sampling is unnecessary or inappropriate. For example, in exploratory studies, the researcher’s goal is to obtain ideas, new insights and experienced critical appraisals just to help him pose a research problem or hypothesis.
The researcher conducting such studies does not carry out the studies of samples with the purpose of being able to generalize to the populations that are being sampled. Thus, he selects a purposive sample.
The respondents are selected precisely because of their special experience, exposures and competence, the market- researchers are typically content with accidental or purposive samples, which are selected in such a way that the likelihood of difference among the elements in the sample is maximized.
ADVERTISEMENTS:
They may be looking for ideas to be transmitted, say, to the persons in charge of advertising the products rather than putting across correct estimates of the population distribution.
Sometimes there is no way out but to resort to the non-probability sampling. If one is trying to find out something, for example, about the experiences of people who had to leave, say, Sri Lanka, owing to certain political developments, he has no realistic choice but to rely on informants who happen to be available, here and now.
Of course, the researcher’s choice here is between data that do not permit a statistical assessment of the margin of error, etc. and no data at all. This does not mean, of course, that one is not concerned with the possibility of error; it is only that he places his reliance on internal consistency of data and its coherence with other information that he might have come to secure.
We must remember that there are many important considerations in research, in addition to the sampling design. It may, therefore, be necessary to balance one consideration against another. Sometimes, wisdom lies in a better and more accurate sampling design being given up in favour of more sensitive method of data collection.
It is in this light that we should understand why the use of non-probability sampling may on occasions be justified. Of course, the decision whether it would be better to gather more adequate or in-depth information based on a not very sound sample or less adequate information based on a sounder sample, is by no means an easy one to arrive at.
It is in terms of the research-purpose that the researcher may take such a decision.
For example, in a study of factors related to the use of narcotics by boys in juvenile street gangs, Chein and Associates (1957) used a sample of social group-workers who had spent quite sometime winning the confidence of gang boys.
This sample was an accidental sample of group-workers and since they could give information about only those gang boys with whom they had worked, the sample of gang members about whom information could be had was also an accidental sample.
But considering the facility with which more dependable information about the gang could be had from these group- workers, the researchers preferred an accidental (non-probability) sample to probability sample of gang members (assuming that it was possible to get such probability sample).
Thus, in his scientific wisdom, the researcher has to carefully weigh the gains and liabilities of various research procedures. He may, under certain circumstances, sacrifice the probability principle in his sampling procedure to gain a deeper understanding through more sensitive and dependable instruments of securing information.