BibTeX
@MISC{Baxter_thelikelihood,
author = {Rohan A. Baxter},
title = {The Likelihood Principle and MML Estimators},
year = {}
}
Share
 |
 |
 |
 |
||
OpenURL
Abstract
: The minimum message length (MML) family of estimators are discrete Bayesian point estimators which violate the Likelihood Principle. The nature of the breach is examined and examples are given. MML proponents have claimed that the violation is an `innocent' one. This claim is examined. Keywords: MML, Likelihood Principle, Minimum Encoding Length Estimators, Bayesian Point Estimation. Area of Interest: Minimum Encoding Length Inference Methods. 1 Introduction In this section, we review the minimum message length (MML) family of estimators. We review three members of this family: Strict MML (SMML) estimators and two approximate estimators to SMML estimators, Fairly Strict MML (FSMML) and MML. We then discuss some interpretations of the `Likelihood Principle' (LP) and describe the interpretation used in this paper. We then show that the MML estimators breach our interpretation of the LP. In sections 2 and 3, examples are used to investigate the nature of the breaches by their effects...
