Models for the association between ordinal variables

Abstract

In this paper we present marginal association and regression models as an alternative to classical association models for cross-classified ordinal data. It is shown that the methods easily incorporate various types of association structures, are able to include covariate information and generalize easily to multi-way classifications. The proposed approach is used to analyze data from a multicentre psychiatric study. (C) 1998 Elsevier Science B.V. All rights reserved.