ABSTRACT
The survey sampler does not always have a list of all the members of a population. In cluster sampling, the population is organized in groups called clusters; clusters, rather than individuals, are selected at the first stage of sampling. Clusters often occur naturally; examples include persons in the same household, residents of the same county, birds in the same nest, or patients in the same hospital. This chapter shows how to use the SURVEYMEANS and SURVEYFREQ procedures in SAS® software to compute estimates from one-stage and multi-stage cluster samples, with code and output for the examples from Chapter 5 of Sampling: Design and Analysis, Third Edition. A macro presented with the book allows students to explore what happens to the actual coverage probabilities of nominally 95 percent confidence intervals when clustering is ignored.
