TESTS OF INDEPENDENCE ON 2-WAY TABLES UNDER CLUSTER SAMPLING - AN EVALUATION

This paper describes a Monte Carlo study of competing procedures for t esting row-column independence in a two-way table, under cluster sampl ing, The study was designed to explore the effects on test performance of a number of parameters, including: (i) different row and column de sign effects; (ii) a range of values of the generalized design effects corresponding to the independence test; (iii) differing degrees of va riation among the generalized design effects, A model of cluster sampl ing was constructed that allows these study parameters to be controlle d, One aim of the study was to explore the extent to which the finding s of Thomas & Rao (1987) for simple goodness-of-fit tests generalize t o the independence test, A second aim of the study was to examine a nu mber of test types not considered by Thomas & Rao (1987), A basic conc lusion of the paper is that practitioners Rave a wide range of viable procedures from which to chose, depending on the nature of their study sample, Some of the issues to be considered by practitioners in selec ting appropriate test statistics are discussed in the paper.