Customer Reviews

An Introduction to Generalized Linear Models, Second Edition

5 star:		(7)
4 star:		(3)
3 star:		(0)
2 star:		(0)
1 star:		(0)

 

 

Bill recommended Dobson's text because of her clear writing style and many useful examples. Dobson also places the theory in the context of the general exponential family of distributions. As I knew that the second edition was about to come out I waited for it.

The wait seems to have been very worthwhile. The second edition is a real bargin.... She has updated it with the many advances that have occurred over the past 12 years since the first edition was printed. This edition now includes some discussion of generalized additive models, broader coverage of applications as survival analysis, GEE, multi-level models and nominal and ordinal logistic regression have been added. It now offers the reader more applications in a wider variety of disciplines and includes modern approaches to diagnostic checking of the models.

As with the first edition, exploratory techniques are emphasized particularly graphical methods. The goal is to unify the apparently disparate statistical techniques that students are exposed to, into one general modeling framework.

It includes a nice up-to-date bibliography and recent advanced results on longitudinal models. The level is intermediate statistics with introductory statistics and linear models taken to be prerequisites. Students are also required to have some familiarity with calculus and linear algebra.

This book provides a surprisingly brief and gentle, yet thorough, introduction to the subject of modeling dependent variables that are not continuous (see note below). The reader, who should be familiar with calculus-based probability, may initially find it frustrating that the actual practice of modeling nominal data is not discussed until the last two chapters (of 9). However, the cause for delaying the discussion of these models is to introduce the terminology and methodology of generalized linear models through more familiar linear regression models.

Thus, while this book is not ideal for someone who wants to jump right into the thick of building logistic, loglinear, or other models for nominal data, it is quite suitable for those wishing a thorough introduction to the practice of generalized linear modeling. For greater detail, a thicker book like McCullagh & Nelder's _Generalized Linear Models_ would be suitable.

Note: While the term "Generalized Linear Models" includes linear regression models (i.e., models for continuous dependent variables), reading this book is not the easiest way to be introduced to regression. A better starting point would be Draper & Smith's _Applied Regression Analysis_ or Weisberg's _Applied Linear Regression_.

I wish somebody would write a concise tutorial of the matematics required for an "intermediate" book such as Dobson's. Undoubtedly for someone whose acquaintence with modern statitical methods is more current this book is a gem. For someone like myself who wants a refresher and whose math is a bit rusty it leaves something to be desired. Some of the theoretical derivations in chapters 3 and 4 (keys to the understanding of the rest of the book) would be improved by a bit more detail and a thoroughly worked example. A major shortcoming is the lack of answers to the excercises; I don't see how the book was published without them. If the book was 100 pages longer with the addition of the aforementioned material, I would have given it a five star rating.

This book does exactly what it set out to do. It was recommended to me as as an excellent introduction to GLMs and in this it succeeds.

Even though it's not stated the book really assumes a knowledge of regression and basic ANOVA. If you havn't a reasonable knowledge of the basics of these, this book is not for you.
Armed with basic knowledge Annette Dobson's book is really good. The background theory is covered in the first 5 chapters. This is well structured and deals with the subject in a sensible manner and at a relatively quick pace. As such, it is ideally suited to the intermediate audience of a senior level lecture course and also to researchers who wish to quickly understand and use GLMs.

The second part of the book focuses on applications and interpretations with some more theory - this overlaps and uses the work of the earlier chapters. The key material is covered and the author quickly explains what the results mean and how they should be interpreted. Once again the exposition is thorough but brief and so it suited to a course work environment or the researcher doing self-study/refreshing their knowledge, but it's not for the novice or those starting out in statistical modelling.

If you are one of those people that like to learn few things and be able to apply them to many, this is a book for you. It provides derivations for properties of a whole family of distributions, which can be applied to each of the member distributions. It is short, sweet, and straight to the point. Basic knowledge of linear algebra and multivariate calculus might be necessary. As a complementary text and for a more detailed discussion, I would also recommend Statistical Models by David Freedman.

The book of Dobson and Barnett is an excellent tool for practitioners in Biostatistics, where They're dealing with non normal variables such as the ones associates to proportions, counts and positive data. The text is both theoric and applied, with several examples.
The third edition has been updated in Bayesian Analysis, which extends the book out of the frequentist perspective. Additionally, the text is presented for general use and applications because the use of R-package, a well known free software.

While what the book does explain about the statistical theory mentioned, it is too compact for what it tries to explain. There are also no answers to the excercises, which would be quite helpful given some of the questions asked. It's great for applications and is a good handbook, but for a thorough explanation of everything involved, I recomend getting a bigger textbook! For my 4th year Generalized Linear Models stats class, this book is helpful, but at times too compact to be more useful.

My copy of the second edition just arrived yesterday and it is even better than the first edition (which was fantastic). The logical organization and clarity of writing make this book a 'must have' for any statistician's library. I'd give it 6 stars if I could. Readers should also check out McCulloch and Searle's 'Generalized, Linear and Mixed Models'.

This book is a great overview for someone who already has taken a course in mathematical statistics with a strong focus on inference and estimation. However, don't expect too much detail on any of the methods. The real strength of this book is that fact that it is concise, and gives you the grand picture of GLMs in a very clear way. If you're looking for a book that goes into more detail but covers less material, I would highly suggest Freedman's Statistical Models: Theory and Practice.

What more needs to be said? This book gives the best accessible introduction to generalized linear models.