Finite-sample analytic properties of percentile bootstrap intervals

被引：0
|
作者：
Wang, Weizhen [1 ,2 ]
Yu, Chongxiu [1 ]
Zhang, Zhongzhan [1 ]
机构：
[1] Beijing Univ Technol, Sch Math Stat & Mech, Beijing 100124, Peoples R China
[2] WRIGHT STATE UNIV, Dept Math ad Stat, DAYTON, OH 45435 USA
来源：
METRIKA | 2025年
基金：
中国国家自然科学基金; 北京市自然科学基金;
关键词：
Binomial distribution; Coverage probability; Expected length; Linear model; Normal distribution; CONFIDENCE-INTERVALS;
D O I：
10.1007/s00184-025-00990-8
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 bootstrap interval is an efficient procedure to estimate parameters. The coverage probability and expected length are crucial to evaluate the reliability and accuracy of a confidence interval. How to compute them for a bootstrap interval at a given parameter configuration for a fixed sample size? In this paper, we offer the first attempt at computing the two quantities of percentile bootstrap intervals by exact probabilistic calculation. This method is applied to ten basic bootstrap intervals for six important parameters. Interestingly, we find that some 1-alpha bootstrap intervals are narrower than the optimal 1-alpha z-interval or t-interval.
引用
收藏
页数：27
相关论文
共 50 条
12345