An AID/Logit Procedure for Analyzing Large Multiway Contingency Tables

A two-step procedure is described in which AID is applied first to a multiway contingency table to isolate those variables (and levels within variable) that best account for variance in the criterion. A logit analysis follows for estimating parameters for prediction purposes. The method is described and applied illustratively to a set of data on usage of an AT&T telecommunications service.