Conditional Specification of Statistical Models

This research monograph will be of interest to mathematical statisticians.

Klappentext

The concept of conditional specification of distributions is not new but, except in normal families, it has not been well developed in the literature. Computational difficulties undoubtedly hindered or discouraged developments in this direction. However, such roadblocks are of dimished importance today. Questions of compatibility of conditional and marginal specifications of distributions are of fundamental importance in modeling scenarios. Models with conditionals in exponential families are particularly tractable and provide useful models in a broad variety of settings.

Zusammenfassung
Efforts to visualize multivariate densities necessarily involve the use of cross-sections, or, equivalently, conditional densities. All statistical researchers seeking more flexible models than those provided by classical models will find conditionally specified distributions of interest.

Inhalt
Conditional Specification: Concepts and Theorems.- Exact and Near Compatibility in Distributions with Finite Support Sets.- Distributions with Normal Conditionals.- Conditionals in Exponential Families.- Other Conditionally Specified Families.- Improper and Nonstandard Models.- Characterizations Involving Conditional Moments.- Multivariate Extensions.- Estimation in Conditionally Specified Models.- Marginal and Conditional Specification in General.- Conditional Survival Models.- Applications to Modeling Bivariate Extremes.- Bayesian Analysis Using Conditionally Specified Models.- Conditional Specification Versus Simultaneous Equation Models.- Paella.