<< 1 >>
Rating: 
Summary: wonderful treatment of outlier issues in regression
Review: These authors provide an excellent guide to the available theory of robust regression. Graphical methods are well used to bring home the points. I have always liked looking at outliers and robustness from an influence function viewpoint and the authors do an excellent job of describing that theory. The book came out in 1987 and so it is a little dated. Another reviewer criticizes some of the software that is mentioned. With computer methodology advancing so rapidly one should not expect the state-of-the-art in computing from a 14 year old text. Value it for its exposition and development of the theory.
Rating:
Summary: wonderful treatment of outlier issues in regression
Review: These authors provide an excellent guide to the available theory of robust regression. Graphical methods are well used to bring home the points. I have always liked looking at outliers and robustness from an influence function viewpoint and the authors do an excellent job of describing that theory. The book came out in 1987 and so it is a little dated. Another reviewer criticizes some of the software that is mentioned. With computer methodology advancing so rapidly one should not expect the state-of-the-art in computing from a 14 year old text. Value it for its exposition and development of the theory.
Rating: 
Summary: Easy to follow introduction
Review: Why should I read this book? - The introduction alone is worth half the price of this book. It gives you a very good overview of robust estimators like m-estimators, least median of squares and least trimmed squares. Basically, you could just take the introduction and start implementing, if that's what you want. The author is able to tell how this and that method overcomes a problem and for what price, and all in one paragraph. What do I need to understand this book? - The math in this book is kept at an engineer's level, so an introductory course on statistics should be enough to follow the introduction and to get an insight into the following chapters. A lot of examples with real life data and nice plots show how the methods perform. What's the drawback? - The author talks a lot about his old-school fortran program called 'PROGRESS' - and eventually comes up with screen dumps (when people were still using terminals) of the program output. He even gives usage information like 'PROGRESS is designed to run on an IBM-PC or a compatible microcomputer. At least 256K RAM must be available.' Neither the source nor the binary is shipped with the book. - Some new trends in robust statistics are not included, in particular, I do miss the RANSAC. Conclusion: - Better get this book from a library and photocopy the introduction. If you're not really interested, the introductory chapter should be enough.
<< 1 >>
| 
