<< 1 >>
Rating: 
Summary: unique practical book on multivariate analysis
Review: Ram Gnanadesikan wrote the first edition of this book in 1977. It was unique then as a practical text on multivariate analysis based on his experience at Bell Labs. There is a nice mix of good theory and practice in the book. Also it is not tied to the theory of the multivariate normal distribution and it emphasizes graphical representations, robust methods, outlier detection and dimensionality reduction techniques. Clustering and classification methods are also covered. Gnandesikan and his colleagues were the first to think of using Hampel's influence function for detecting outliers in multivariate data. Their research is covered in this book. I used it in my work at Oak Ridge National Laboratory in the late 1970s to detect multivariate outliers as part of our energy data validation effort. I also applied these ideas to time series analysis. Twenty years after the publication of the first edition, Gnanadesikan decided to produce the long overdue revision. He is currently retired from Bell Labs and is employed as a Professor of Statistics at Rutgers University. The second edition incorporates advances from the last 20 years and emphasizes the newly available software. The new computer-intensive methods including the bootstrap are not covered.
<< 1 >>
| 
