Methods for analyzing multivariate binary data, with association between outcomes of interest.

A likelihood based method is proposed for multivariate categorical data. It is assumed that, together with the marginal outcomes, the set of pairwise associations between outcomes is of scientific interest. The focus is on binary outcomes and it is indicated how the proposed method generalizes to categorical outcomes. A connection with second-order generalized estimating equations (GEE2) is established. The method is applied to analyze data from a developmental toxicity study.