Partitioning chi-square the analysis of contingency tables with repeated measurements

Abstract

Categorical variables are commonly used in gerontological research. Although analysis of such variables presents no problem as long as the observations are independent, problems do arise when they are found in a design with repeated measurements. In this paper, two procedures from the statistics literature are combined to provide an analysis for data from a mixture of between and within categorical variables. First, a response pattern is used to form a contingency table with observations which are completely independent. Next, the Pearson chi-square statistic from the table is partitioned to provide tests analogous to simple effects in analysis of variance.