Resampling Methods for Dependent Data
Resampling Methods for Dependent Data
复制标题
DOI：
10.1007/978-1-4757-3803-2
复制
发表时间：
2003-08
期刊：
影响因子：
--
通讯作者：
S. Lahiri
中科院分区：
文献类型：
--
作者：
S. Lahiri
这是一本关于时间和空间数据的Bootstrap和相关重采样方法的书，这些数据表现出各种形式的相关性。与独立数据的重抽样方法一样，这些方法提供了对相关数据进行统计分析的工具，而不需要严格的结构假设。这是相依情况下重抽样方法的一个重要方面，因为模型错误说明的问题在相依情况下更为严重，而传统的统计方法往往对偏离模型假设非常敏感。随着Efron(1979)的Bootstrap方法在解决许多涉及独立数据的复杂问题上取得了巨大的成功，并遵循Singh(1981)关于依赖下方法的不充分性的例子，文献中已经有几次尝试将Bootstrap方法推广到依赖情况。由Hall(1985)、Carlstein(1986)、Kiinsch(1989)、Liu和Singh(1992)等人以不同的形式和不同的推理问题提出的分段重采样取代了单一观测的重采样，取得了突破性进展。从那时起，相依数据的处理方法有了长足的发展，至今仍是一个活跃的研究领域。这本书描述了理论和方法的各个方面的重采样方法的从属数据发展在过去二十年。这本书主要有两个目标受众，相关部分的阐述层次根据每个受众而量身定做。
This is a book on bootstrap and related resampling methods for temporal and spatial data exhibiting various forms of dependence. Like the resam pling methods for independent data, these methods provide tools for sta tistical analysis of dependent data without requiring stringent structural assumptions. This is an important aspect of the resampling methods in the dependent case, as the problem of model misspecification is more preva lent under dependence and traditional statistical methods are often very sensitive to deviations from model assumptions. Following the tremendous success of Efron's (1979) bootstrap to provide answers to many complex problems involving independent data and following Singh's (1981) example on the inadequacy of the method under dependence, there have been several attempts in the literature to extend the bootstrap method to the dependent case. A breakthrough was achieved when resampling of single observations was replaced with block resampling, an idea that was put forward by Hall (1985), Carlstein (1986), Kiinsch (1989), Liu and Singh (1992), and others in various forms and in different inference problems. There has been a vig orous development in the area of res amp ling methods for dependent data since then and it is still an area of active research. This book describes various aspects of the theory and methodology of resampling methods for dependent data developed over the last two decades. There are mainly two target audiences for the book, with the level of exposition of the relevant parts tailored to each audience.