Bayesian hypotheses testing using posterior density ratios

被引：9

作者：

Basu, S ^{[1
]}

机构：

[1] UNIV ARKANSAS,DEPT MATH SCI,FAYETTEVILLE,AR 72701

来源：

STATISTICS & PROBABILITY LETTERS | 1996年 / 30卷 / 01期

关键词：

Bayes factor; hypothesis test; likelihood ratio; posterior; robust Bayes;

D O I：

10.1016/0167-7152(95)00204-9

中图分类号：

O21 [概率论与数理统计]; C8 [统计学];

学科分类号：

020208 ; 070103 ; 0714 ;

摘要：

The posterior density ratio sup(theta is an element of H0) pi(theta/data)/sup(theta is an element of H1) pi(theta/data) is proposed as a new criterion for Bayesian hypothesis testing. The performances of the posterior density ratio and the posterior probability of H-0 are compared in point null hypothesis, precise hypothesis and one sided hypothesis tests. It is found that the likelihood ratio and the lower bound of the posterior density ratio over a scale mixture class of priors match in point null hypothesis testing. They, however, do not match in one sided testing.

引用

页码：79 / 86

页数：8

共 12 条

[1]

BASU S, 1992, IN PRESS SANKHYA A

[2]

BERGER J, 1994, TEST, V2

[3]

Berger J. O., 1987, STAT SCI, V2, P317

[4]

BERGER JO, 1987, J AM STAT ASSOC, V82, P112, DOI 10.2307/2289131