摘要
从1979年Efron提出Bootstrap方法至今,该方法在近30年间已经得到了极大的发展和扩充,并被广泛地应用于统计学的各个领域。Bootstrap理论的基本思想、历史发展及其若干比较前沿的研究方向包括:独立同分布数据、基于模型、带有块结构、Sieve、基于变换、Markov过程、长期相依和空间数据的Bootstrap理论,其中对独立同分布数据的Bootstrap应用最为基础,其余七种方向都可以视为附加了各种特殊条件的Bootstrap应用。由于Bootstrap的应用通常需要一定的统计程序编写,在介绍各种研究方向的同时,也相应简要介绍一些算法实现,其软件工具采用当今国际统计研究的主流语言——R语言。
Bootstrap methods have been developed and extended in a large number of fields of statistics ever since the first inception by Efron （1979）. This paper gives an introduction of the basic methodology, historical developments and some frontiers of the bootstrap theory. The main directions introduced in this paper includes： bootstrap for i. i. d data, model - based bootstrap, block bootstrap, sieve bootstrap, transform- based bootstrap, bootstrap for Markov process, bootstrap for long range dependence and bootstrap for spatial data, among which the first direction is the most fundamental, and other directions can be regarded as extensions under certain conditions. At the same time, we＇ 11 give some program scripts for different bootstrap algorithms; our tool is R language, which is being widely adopted nowadays in implementing new methods and models of statistics by international academicians.
出处
《统计与信息论坛》
CSSCI
2008年第2期90-96,共7页
Journal of Statistics and Information
