Analysis of Variance for Random Models, Volume II: Unbalanced Data, Theory, Methods, Applications, and Data Analysis
Analysis of Variance for Random Models, Volume II: Unbalanced Data, Theory, Methods, Applications, and Data Analysis
复制标题
随机模型的方差分析，第二卷：不平衡数据、理论、方法、应用和数据分析
DOI：
10.1198/tech.2006.s350
复制
发表时间：
2006
期刊：
影响因子：
2.5
通讯作者：
Thomas R. Gatliffe
中科院分区：
文献类型：
--
作者：
Thomas R. Gatliffe
我们许多每天处理微量环境污染物数据的人都急切地期待着对左删减测量数据的真正全面处理。虽然大多数实践者都有必要在一定程度上适应“不可靠地可检测”或“不可测量”的观察报告的处理，但Dennis Helsel在这本书中提供了对当前替代方法谱的现成参考，并充分记录了它们在各种代表性环境和数据质量条件下的适当使用。主题以非常可读的散文呈现，并以易于遵循的数学复杂性发展。许多例子来自于以前发表的关于空气、水、土壤和生物群质量监测的工作，并说明了采用各种方法处理“未检测到”的优点和缺点。在其12个章节中，本书相对侧重于在数据分析中使用“非检测”观测的三种根本不同的方法：(1)用固定或可变值替代非检测值，(2)潜在分布参数的最大似然估计，以及(3)部分或全部非参数方法。在一系列统计问题上，包括计算汇总统计和区间估计、比较组等价性、确定数据组相关性以及用回归和趋势分析进行预测，探讨了每种方法的局限性和最佳应用。本文探讨了与检测、定量和定向报告限制相关的常见的主观和客观信息审查类型的影响，并详细讨论了每种类型所带来的特殊问题。比较了不同分析中由于数据中存在单个或多个审查水平而引起的并发症以及适当的补偿措施。低于报告限额的所有数据的特殊情况，以及可以适当地推导和应用哪些统计数据和检验的问题，都有单独的一章。为了帮助说明书中提出的技术，每章都以两到四个基于提供的现实生活数据集的正式练习结束。Helsel博士还维护一个网站，提供Minitab®、MS Excel®和R格式的各种数据集的电子版本，以增加和补充文本；课文练习的解决方案；和Minitab®版本14宏专门开发来实现所讨论的方法。虽然文本在其示例中使用Minitab®软件，但所有技术和算法都进行了足够详细的讨论，以使其能够与几乎同等复杂或更好的几乎任何其他商业软件包成功实现。例如，在阅读文本和准备这篇综述时，我能够相对容易地编写用于S-PLUS®和Systat®的几种技术的脚本。然而，我也发现很多材料都可以很容易地适应使用，仅仅是商业电子表格软件（如MS Excel®）提供的补充统计功能。因为它只关注环境数据分析的一个相当狭窄的方面，所以它可能最好被用作学期/季度课程的补充参考资料，或者作为Helsel博士通过他的实用统计网站提供的一到两天的短期课程的主要参考资料。作为个人参考书的补充，我将把这本书推荐给任何进行或审查环境数据分析和报告的人。
Many of us who work with trace-level environmental contaminant data on a daily basis have eagerly awaited a truly comprehensive treatment of the statistics for left-censored measurements. Although most practitioners have of necessity developed some level of accommodation with the treatment of observations reported as “less than reliably detectable” or “immeasurable,” with this book Dennis Helsel provides both a ready reference to the current spectrum of alternative methods and a fully documented discussion of their appropriate use in a variety of representative circumstances and conditions of data quality. The subject matter is presented in very readable prose and developed with easily followed mathematical sophistication. Many of the examples are drawn from previously published work in air, water, soil, and biota quality monitoring, and illustrate both the advantages and the shortcomings of employing the various methodologies for handling “nondetects.” Throughout its 12 chapters the book comparatively focuses on the three fundamentally differing approaches for using “nondetect” observations in data analysis: (1) Substitution of fixed or variable values for nondetects, (2) maximum likelihood estimation of underlying distribution parameters, and (3) partially or wholly nonparametric methods. The limitations and optimal applications for each are explored for a range of statistical questions, including computing summary statistics and interval estimates, comparing group equivalence, determining data group correlation, and prediction with regression and trend analysis. The effects of commonly encountered types of subjective and objective information censoring relative to detection, quantitation, and directed reporting limits are explored, and the special problems posed with each are discussed in detail. The complications caused by the existence of single or multiple censoring levels in the data and appropriate compensatory measures in different analyses are compared. The special case of all data below the reporting limit and the question of which statistics and tests may be appropriately derived and applied are given their own chapter. To help illustrate the techniques presented in the book, each chapter concludes with two to four formal exercises based on supplied real-life data sets. Dr. Helsel also maintains a website with material to augment and supplement the text, providing electronic versions of the various data sets formatted for Minitab®, MS Excel®, and R; worked solutions for the text exercises; and Minitab® version 14 macros specifically developed to implement the discussed methods. Although the text employs Minitab® software in its examples, all techniques and algorithms are discussed in sufficient detail to allow their successful implementation with almost any other commercial package of nearly equivalent sophistication or better. For example, while reading the text and preparing this review, I was able to relatively easily script several of the techniques for use with S–PLUS® and Systat®. However, I also found that much of the material can be easily adapted for use with nothing more than the supplemental statistical function capabilities available with commercial spreadsheet software such as MS Excel®. Because it treats a fairly narrowly focused aspect of environmental data analysis, the text would probably be best employed as a supplemental reference source within a semester/quarter-length course or as a primary reference for a oneor two-day short course such as offered by Dr. Helsel through his Practical Stats website. As a prospective addition to a personal reference library, I would recommend this book for anyone conducting or reviewing the analysis and reporting of environmental data containing significant left-censoring.
DOI：
10.1109/tpami.1984.4767596
发表时间：
1984-01-01
影响因子：
23.6
作者：
GEMAN, S;GEMAN, D
通讯作者：
GEMAN, D