Outliers in Statistical Data (Wiley Series in Probability & Statistics)

Rating:
Summary: thorough treatment of the subject of outliers
Review: The first edition of this book in 1978 brought the first comprehensive text on outlier methods in statistical analysis. Outliers present important practical problems that have been studied for centuries. The twentieth century brought appropriate formal statistical methods for the detection and treatment of outliers.

The literature on outlier methods was scattered. It is commonly discussed in texts on regression analysis and sometimes in texts on robust statistical methods. Its importance in time series analysis was overlooked for a long time. Gnandesikan took the novel approach of using influence functions to search for multivariate outliers and covered this in his multivariate text. Another monograph exclusively on outlier methods by Doug Hawkins was also published at that time. However Barnett and Lewis gave the first comprehensive and unified account of the existing literature.

The second edition published in 1984 provided an update and added chapters on outliers in spatial data and outliers in time series. The chapter on time series devoted some space to my approach for using an influence function matrix to detect outliers of unspecified type. This third edition that I am reviewing, published in 1995 expanded on the work in the previous editions, restructured the text and added many new references. It continues to be the classic reference on outlier methods.