Admissibility and Complete Classes

  • Christian P. Robert
Part of the Springer Texts in Statistics book series (STS)

Abstract

The previous chapters mentioned repeatedly that the Bayes estimators were instrumental for the frequentist notions of optimality, in particular, for admissibility. This chapter provides a more detailed description of this phenomenon. In §6.1, it considers the performances of the Bayes and generalized Bayes estimators in terms of admissibility. Then, §6.2 studies Stein’s sufficient condition in order to relate the admissibility of a given estimator with a sequence of prior distributions. The notion of complete class introduced in §6.3 is also fundamental, as it provides a characterization of admissible estimators or at least a substantial reduction in the class of acceptable estimators. We show that, in many cases, the set of the Bayes estimators constitutes a complete class and that, in other cases, it is necessary to include generalized Bayes estimators. In a more general although non-Bayesian perspective, §6.4 presents a method introduced by Brown (1971) and developed by Hwang (1982b), which provides necessary admissibility conditions. For a more technical survey of these topics, see Rukhin (1994).

Keywords

Prior Distribution Exponential Family Admissibility Condition Quadratic Loss Continuous Risk 
These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

Preview

Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.

Copyright information

© Springer Science+Business Media New York 1994

Authors and Affiliations

  • Christian P. Robert
    • 1
  1. 1.URA CNRS 1378 — Dépt. de Math.Université de RouenMont Saint Aignan CedexFrance

Personalised recommendations