Generalized Linear Models With Examples in R (Springer Texts in Statistics)

Five stars for a well-paced, thorough, and practical book for learning GLMs.Its more of an applied book, filled with practical examples, that covers a good bit of theory. The theory is presented at a level geared more towards understanding what's going on "under the hood" than in providing solid mathematical derivations. I liked it for that reason--I was able to feel comfortable in applying GLMs real-life settings rather quickly. I felt I had the tools to ask the right questions and have an idea of "knowing what I didn't know." Its a very good starter book, and you will come out understanding the common applications of the core glm models. However, you'll need to get more detailed resources if need to go beyond that.About a 3rd of the book is devoted to the applying the linear model which helps build background and intuition for GLMs. These chapters are organized into a rough order that can be described as model theory, estimation, inference, model comparison/evaluation, and diagnostics. The chapters on GLMs follow the same outline so parallels between the two are very clear. In this way the key ideas used in modeling, whether linear models or glms, are emphasized nicely.The linear model chapters and GLM chapters is are bridged by a good discussion of maximum likelihood. At least I came out with an appreciation of the connection between least-squares, iteratively weighted least squares and maximum likelihood estimation. The chapters on GLMs cover the exponential dispersion family, dispersion, and variance functions. The authors connect this to interpretations of transforms in linear models and to the different types of residuals, their uses and interpretations. The authors are also then able to talk about the quasi-binomial (and other quasi models etc.) , which was very interesting.The core GLM chapters are followed by individual chapters focusing applying specific models to specific types of data. These are the common applications, gamma for positive continuous, binomial for porportions etc. These chapters were the weakest in my opinion, mostly because they felt rushed (not rushed in writing, but rushed in that I wish they could cover more). The chapter on Poisson glms was the weakest for me, mostly because I've never encountered contingency tables, and I didn't feel like that chapter gave me enough understanding to "ask the right questions" if I had to deal with modeling one.What really makes the book work for me is that they are actually doing stuff using R from the get-go. You get to 'handle' the things you are talking about immediately, and you can practice R by just following along and typing their examples. A really nice feature of the book is that there is an index of R functions, and most applied chapters have an ending summary section that reviews the key commands used in that chapter. This is pretty handy, and I find that their descriptions are sometimes much clearer than the R documentation itself. All the data sets come with an R package and there is a solution set in the back to some problems. Its not perfect, but I got a lot out of it.I feel the authors thoroughly motivate and explain the theory underlying GLMs as well as how to practically handle them, and to interpret results in terms of theory. At times technical explanations are terse, but a book like this can't let "technical details" overtake understanding and practice of applications. That's not to say they brush over the theory - they go pretty deep and you can get a very advanced understanding of it.

There are not many modern textbooks on this topic, so I was delighted to have found this book. It is clear and well written. The first part of the book provides coverage on general linear models, and afterwards moves onto the theory of GLMs. After covering the theory, applications are then introduced. This book differs from some others on the same topic due to its coverage of tweedie GLMs (an area of research of the authors).Overall, this book provides a clear and concise coverage of an important topic in statistics. The R code is also very helpful to apply the methods.

This is a fantastic book for anyone looking to understand generalized linear models (GLM). I think it is much better than Dobson & Barnett's  An Introduction to Generalized Linear Models  (which is still a great book itself), and it is certainly easier to understand than McCullagh & Nelder's seminal text  Generalized Linear Models  (since the latter is more of a theory book). I strongly recommend it to anyone working in (applied) statistics, data analysis/science, and related areas.I only have positive things to say about the book. There are plenty of R code examples and figures, as well as descriptions of what different R functions do. There are several "case studies" throughout the text which help reinforce the concepts introduced in each chapter. The authors also provide an R package, "GLMsData", specifically for data analysis practice on real datasets. Additionally, the authors also created the "tweedie" and "statmod" R packages. There is an Appendix with a basic introduction to using R, and there is a separate Index for datasets, R commands, and general topics/keywords.Chapters 1 to 3 are introductory. Chapter 2 is a good overview of (standard/normal) linear models, including model/parameter estimation, inference and testing, ANOVA, and model comparison/selection. Chapter 3 covers linear model assumptions and diagnostics, such as leverage, residuals, and outliers/influential observations. Importantly, they also describe variable transformations (including Box-Cox), polynomial trends, and a brief section on splines.Chapter 4 is the first to go "beyond linear regression". The focus of the chapter is (maximum) likelihood estimation, and includes Fisher scoring, testing, and confidence intervals for non-normal outcomes. This chapter will likely be challenging unless you have a strong linear algebra background.Chapters 5-8 are the "heart" of the book: * Chapter 5 is the first introduction to GLM *structure* (random and systematic components) and introduces "exponential dispersion models (EDM)". This chapter includes two of the most useful tables I have seen: Table 5.1 shows common EDMs along with their variance functions, canonical and dispersion parameters, etc. Table 5.2 lists variance-stabilizing transformations and the GLMs they approximate. * Chapter 6 covers parameter estimation (including dispersion) and is relatively short. Table 6.4 is extremely useful; it lists the names of link functions in R that are listed by the different "glm()" families. * Chapter 7 covers model inference, hypothesis testing, model comparison, and analysis of deviance. * Chapter 8 is the GLM diagnostics chapter. Helpfully, the authors list the important assumptions made when fitting GLMs, and discuss most of what was covered in Chapter 3 for standard LMs. The authors also introduce *quantile residuals* (calculated with their "statmod" package) as well as quasi-likelihood.Chapter 9-12 cover specific GLMs, with each chapter including a bit of theory but mostly applied/data analysis material. Each chapter is for a different type of *outcome/response variable*: * Chapter 9 covers proportions (binomial GLMs) * Chapter 10 covers count data (Poisson and negative binomial GLMs) * Chapter 11 covers positive continuous data (Gamma and inverse gamma GLMs) * Chapter 12 might be the most valuable, discussiong *Tweedie GLMs*. They describe how the Tweedie family of EDMs relate to the other EDMs in the book (connected by the form of their variance functions). The chapter ends with some case studies and examples using functions from their "tweedie" and "statmod" R packages.In summary, I think this book is a "must-have". If you struggled to fully understand  McCullagh & Nelder  and felt that  Dobson & Barnett  was not in-depth enough, then you will probably enjoy this book. While most of the material is appropriate for beginners, there is still more advanced material that require a good understanding of linear algebra and, to a lesser extent, calculus.

This is by far the nastiest little book I have ever purchased at this price. It is bound so tightly that it can’t stay open except crossing p 250. You have to hold the animal with one hand or purchase a heavy smartphone to keep the book open but it will remain a struggle to see the text near the binding. The text fills the pages with small margins and the paper is glossy and wavy. I have finished the first three chapters and I hope I can reach the end without sustaining an injury. You clearly see that the thing is meant to secure high returns to the stakeholders. This is very unfortunate because the book - I mean the content - is good. I would rate the book highly if it had not been for the poor job at Springer.

Ich habe mir das Buch als Begleitung zu einem sehr mathematischen Kurs in GLM gekauft. Die Mathematik kommt für meine Verhältnisse zu kurz, aber dafür hab ich mir das Buch auch nicht gekauft.Das Buch bietet so viele R Beispiele die allesamt kommentiert sind, und weil nicht zusammenhängend leicht mutprogrammiert werden können. Weiters finden sich viel Übungen wo man das gelernt auch gleich ausprobieren kann! Definitive Kaufempfehlung von mir - auch wenn wie gesagt die Mathematik viel zu kurz kommt, aber wen das nicht stört - TOP