Customer Reviews

Generalized Additive Models (Chapman & Hall/CRC Monographs on Statistics & Applied Probability)

5 star:		(3)
4 star:		(0)
3 star:		(0)
2 star:		(0)
1 star:		(0)

 

 

Believe it or not this important topic in applied statistics was developed by Hastie and Tibshirani when they were graduate students, a sure sign of greatness to come. After their graduation this highly acclaimed book came out explaining both the theory and pratice of generalized additive models in a clear and concise way. Generalized additive models are similar to generalized linear models in their additive structure but the form of the additive functions is much more general.
Both Hastie and Tibshirani are now Stanford professors in the Statistics Department and both have written other excellent books including their joint publication with Jerry Friedman "The Elements of Statistical Learning" and Tibshirani along with Efron wrote an excellent monograph on bootstrap.

A medical doctor recommended that I (a statistician) use generalized additive models on a project that he was involved with. We were pleased with the results of the modelling as they gave us some new insights on his topic and it was very rewarding

Because we were so impressed with the results, we both took the above book out of the library to read. I started the book with some trepidation as I feared it was going to be a complex topic and in some respects I was right. However, right from the encouraging quotes in the preface and into the text itself, I felt the authors were making a great effort to make the book readable.

The medico found that the explanations and examples were well written. His comment was that even a non-statistician could grasp what was going on but that there was enough theory that a statistician would be happy too!

Generalized Additive Models (GAMs) represent the modern and computational extension of Generalized Linear Models (GLMs) (see McCullagh and Nelder) which allow for nonlinear effects in the predictor variables. These nonlinear effects are specified nonparametrically through estimation using a scatterplot smoother. GAMs are frequently used in industry (e.g. FICO scores) due to their specification flexibility and ability to provide a means of dimension reduction. This book provides an introduction to Generalized Additive Models straight from the source. There is not really another book on the subject worth considering.