Abstract
A unified approach to the analysis of categorical data is descussed. The procedure is appropriate whenever one set of categorical variables can be regarded as dependent and a second set is identified as predictor variables. The approach is identical to the tradional logit analysis model, but for computational convenience the procedure can be modeled in the more general log-linear framework. An evaluation of a court-ordered modification of a personnel selection process is used to illustrate the application of the procedure.