Laplace approximations to hypergeometric functions of two matrix arguments

被引：8

作者：

Butler, RW

Wood, ATA ^{[1
]}

机构：

[1] Univ Nottingham, Sch Math Sci, Nottingham NG7 2RD, England

[2] Colorado State Univ, Dept Stat, Ft Collins, CO 80523 USA

来源：

JOURNAL OF MULTIVARIATE ANALYSIS | 2005年 / 94卷 / 01期

基金：

美国国家科学基金会;

关键词：

asymptotic approximation; eigenvalue distribution; matrix-argument hypergeometric function; laplace approximation;

D O I：

10.1016/j.jmva.2004.05.010

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：

We present a unified approach to Laplace approximation of hypergeometric functions with two matrix arguments. The general form of the approximation is designed to exploit the Laplace approximations to hypergeometric functions of a single matrix argument presented in Butler and Wood (Ann. Statist. 30 (2002) 1155, Laplace approximations to Bessel functions of matrix argument, J. Comput. Appl. Math. 155 (2003) 359) which have proved to be very accurate in a variety of settings. All but one of the approximations presented here appear to be new. Numerical accuracy is investigated in a number of statistical applications. (c) 2004 Elsevier Inc. All rights reserved.

引用

页码：1 / 18

页数：18

共 22 条

[1] Laplace approximation for Bessel functions of matrix argument [J].