Conditionally specified distributions (Q1189574)

The focus of the monograph is the study of general classes of conditionally specified distributions. Even though the concept of conditional specification is not new, there is need for the development of inferential techniques for the analysis of data following conditionally specified models. The monograph provides a preliminary guide to such models. The material is covered in 11 chapters. A survey of basic results on the characterization of compatible marginals and the uniqueness of the corresponding joint distribution is given in one of the first chapters for a variety of settings. The subsequent chapters include (i) the development of normal conditional models, (ii) conditionals in prescribed exponential families, (iii) conditionally specified distributions not involving exponential families, (iv) utility of certain improper conditionally specified models, (v) distributional characterizations involving a mixture of regression assumptions and conditional specifications, (vi) extension to higher dimensions, (vii) estimation techniques for conditionally specified models and (viii) methods of simulating data sets from conditionally specified distributions.