On applications of marginal models for categorical data ## AbstractThe paper considers marginal models for categorical data and after reviewing the most important theoretical results concerning the definition, estimation and testing of such models, discusses a number of common statistical problems. These examples include, among others, the analysis of repeated measurements, panel studies and missing data. Fitting marginal models in these cases has the potential of providing the researcher with substantial new insight. The examples illustrate that the marginal modeling approach may be used more widely than thought before. One of the examples shows howgraphical models associated with directed acyclic graphs can be parameterized. A general algorithm is presented to compute maximum likelihood estimates under marginal models
[Edit] |