ON THE INCONSISTENCY OF BOOTSTRAP DISTRIBUTION ESTIMATORS

被引：13

作者：

HALL, P

HARDLE, W

SIMAR, L

机构：

[1] AUSTRALIAN NATL UNIV,CANBERRA,ACT 2601,AUSTRALIA

[2] UNIV CATHOLIQUE LOUVAIN,CORE,B-1348 LOUVAIN,BELGIUM

[3] FAC UNIV ST LOUIS,B-1000 BRUSSELS,BELGIUM

来源：

COMPUTATIONAL STATISTICS & DATA ANALYSIS | 1993年 / 16卷 / 01期

关键词：

BOOTSTRAP; CONSISTENCY; ITERATED BOOTSTRAP; MAXIMUM; RANK; TIE;

D O I：

10.1016/0167-9473(93)90241-K

中图分类号：

TP39 [计算机的应用];

学科分类号：

081203 ; 0835 ;

摘要：

We show that bootstrap distribution estimators of a ranked parameter value are consistent if and only if there are no ties for the rank in question. When inconsistency occurs, the bootstrap distribution estimator does not even converge in probability. This asymptotic result has important implications for small to moderate sample sizes, where poor distribution estimators can result when there are no ties but there are two or more closely spaced parameter values competing to the same rank. Several ways of alleviating the problem of inconsistency are suggested.

引用

页码：11 / 18

页数：8

共 13 条

[1] BOOTSTRAP OF THE MEAN IN THE INFINITE VARIANCE CASE [J].