sampling

Definition:
The process of selecting a subgroup of a population of interest for the collection of information which may be generalized to the whole population, as opposed to a census in which information is collected from the entire population. In marketing research, the population is usually of human beings (but may be of shops, cars etc.). Samples are either probability (or random), or non-probability. Measures of statistical error may be readily applied to probability samples, which, however, require the existence or possibility of a list of all members of the population, and the ability to contact a designated member. The electoral register is often used in the UK as such a list, and calls made at the addresses given. If a minority group is involved, the method is clearly costly. Non-probability samples are selected by researchers or interviewers in accordance either with no rules (purposive) or with rules (quota). The application of rules to sample selection designed to reduce statistical error by making the sample representative of selected subgroups is termed stratification, and the process of concentrating the sample geographically to save travelling for interviews at the expense of increased statistical error is called CLUSTERING. Both are used in complicated surveys to increase efficiency. It should be noted that the statistical computations required to estimate the statistical or sampling errors in these complex samples are themselves complex in calculation, and that the simple formulae appropriate to simple samples will normally underestimate the sampling error.

Cross-References:
[design factor] [random sampling] [clustering techniques]

Links:

Figures: