THE EXPLORATORY ANALYSIS OF SURVEY DATA USING LOG LINEAR-MODELS

The analysis of cross-classified discrete data using log-linear models is now as commonplace as the analysis of continuous data using linear models and yet clear guidance on methods for model selection is scarce. This paper seeks to redress the balance. It focuses on the common situation in which the researcher has no firm a priori model in mind. The general procedures of exploratory data analysis are discussed and the Bayesian Information Criterion is advocated as a useful guide to the most appropriate model. No detailed knowledge of log-linear models is assumed.