Correspondence analysis and one of its generalizations are presented as tools that can be used for the analysis of residuals from log-linear models. By recognizing relations between correspondence analysis and log-linear analysis a better understanding of correspondence analysis is obtained. Furthermore, it is shown how these relations can be used to arrive at a combined approach to contingency table analysis using both log-linear analysis and correspondence analysis.

No abstract

No abstract

We consider a general algorithm for fitting a class of models to tables where the interactions can be described in terms of biplots of rows and columns.