APPROXIMATIONS OF MARGINAL TAIL PROBABILITIES AND INFERENCE FOR SCALAR PARAMETERS

被引：69

作者：

DICICCIO, TJ ^{[1
]}

FIELD, CA ^{[1
]}

FRASER, DAS ^{[1
]}

机构：

[1] YORK UNIV,DEPT MATH,N YORK M3J 1P3,ONTARIO,CANADA

来源：

BIOMETRIKA | 1990年 / 77卷 / 01期

关键词：

Bayesian inference; Conditional inference; Linear regression model; Location-scale model; Lugannani-Rice formula; Saddlepoint approximation; Signed root likelihood ratio statistic; Type II Censoring;

D O I：

10.2307/2336051

中图分类号：

Q [生物科学];

学科分类号：

07 ; 0710 ; 09 ;

摘要：

In many situations, inference for a scalar parameter in the presence of nuisance parameters requires integration of either a joint density of pivotal quantities or a joint posterior density. For such inference, accurate approximations of marginal tail probabilities are useful to avoid high-dimensional integrals. Two tail probability approximations are developed in this paper. Numerical results given for conditional inference in location-scale and linear regression models show the approximations to be generally accurate even for small sample sizes.

引用

页码：77 / 95

页数：19

共 29 条

[1]

BARNDORFFNIELSE.OE, 1988, STAT SCI, V3, P228

[2]

BARNDORFFNIELSEN O, 1979, J ROY STAT SOC B MET, V41, P279

[3]

BARNDORFFNIELSEN OE, 1984, J ROY STAT SOC B MET, V46, P483

[4] ON THE LEVEL-ERROR AFTER BARTLETT ADJUSTMENT OF THE LIKELIHOOD RATIO STATISTIC [J].