An Introduction to Statistical Modeling of Extreme Values
An Introduction to Statistical Modeling of Extreme Values
复制标题
DOI：
10.1198/tech.2002.s73
复制
发表时间：
2002-11
期刊：
影响因子：
2.5
通讯作者：
Eric P. Smith
中科院分区：
文献类型：
--
作者：
Eric P. Smith
极值建模对于水文学、土木工程、环境科学、海洋学和水文学等领域的科学家都具有重要意义。StuartColes关于极值建模的书提供了关于该主题的介绍性文本。它是一个面向建模的文本，重点是不同类型的数据和分析方法。这本书分为九章。在介绍性材料和必要的理论背景讨论之后，是关于极值方法的章节，重点是极值分析中可能使用的不同类型的数据。这些模型包括块最大值模型、阈值模型、来自平稳和非平稳过程的数据模型以及基于点过程的方法。一章涵盖了多变量极值的分析，而第二章布里布里涵盖了贝叶斯推理，马尔可夫链和空间极值等主题。虽然这不是一个数据驱动的文本，但它确实包含了许多例子和分析。这些例子是用来说明方法的;我更希望看到更多的动机和对分析结果的解释。用于文中分析的数据集和S-PLUS程序可在网站上获得。对于那些稍微熟悉S-PLUS的人来说，这些都很容易使用。附录描述了程序，并说明了如何访问数据和使用程序。它还提供到提供其他软件的网站的链接。正文不包括问题集;这些将是有用的，特别是如果正文将在课程中使用。Reiss和托马斯（2001）的文本包含了更彻底的分析数据集，尽管它的长度是正在审查的文本的两倍，而且没有那么精简。这本书是为具有中等统计学背景的个人而写的。那些有最大似然法课程的人应该没有阅读和理解文本的困难。总的来说，这是一个很好的文本，有人开始在极值方法。
The modeling of extreme values is important to scientists in such elds as hydrology, civil engineering, environmental science, oceanography and nance. Stuart Coles’s book on the modeling of extreme values provides an introductory text on the topic. It is a modeling-oriented text with an emphasis on different types of data and analytical approaches. The book is laid out in nine chapters. Following introductory material and discussion of necessary theoretical background are chapters on approaches to extreme values that focus on the different types of data that might be used in an extreme value analysis. These include models for block maximums, threshold models, models for data from stationary and nonstationary processes, and approaches based on point processes. A chapter covers analysis of multivariate extremes, and the nal chapter brie y covers such topics as Bayesian inference, Markov chains, and spatial extremes. Although this is not a data-driven text, it does contain numerous examples and analyses. These examples are used to illustrate the methodology; I would have preferred to see more motivation and interpretation of the results of the analyses. Datasets and S-PLUS programs for the analyses in the text are available at a website. These are easy to use for those slightly familiar with S-PLUS. The appendix describes the programs and illustrates how to access data and use the programs. It also gives links to sites that provide other software. The text does not include problem sets; these would have been useful, especially if the text is to be used in coursework. The text by Reiss and Thomas (2001) contains more thoroughly analyzed datasets, although it is twice the length and not as streamlined as the text under review. The book is meant for individuals with moderate statistical background. Those with coursework in maximum likelihood methods should have no dif culty reading and comprehending the text. Overall, this is a good text for someone getting started in extreme value methods.