cluster sample

cluster sampling employs only limited portions of the population. This may be for a number of reasons – there may not be available a list which effectively defines the population. For example, if an education researcher wished to study 11 year old students, it is unlikely that a list of all 11 year old students would be available. Consequently, the researcher may opt for approaching a number of schools each of which might be expected to have a list of its 11 year old students. Each school would be a cluster.

In populations spread over a substantial geographical area, random sampling is enormously expensive since random sampling maximizes the amount of travel and consequent[Page 26] expense involved. So it is fairly common to employ cluster samples in which the larger geographical area is subdivided into representative clusters or sub-areas. Thus, large towns, small towns and rural areas might be identified as the clusters. In this way, characteristics of the stratified sample may be built in as well as gaining the advantages of reduced geographical dispersion of participants or cases. Much research is only possible because of the use of a limited number of clusters in this way.

In terms of statistical analysis, cluster sampling techniques may affect the conceptual basis of the underlying statistical theory as they cannot be regarded as random samples. Hence, survey researchers sometimes use alternative statistical techniques from the ones common in disciplines such as psychology and related fields.