Customer Reviews

Linear Models with R (Chapman & Hall/CRC Texts in Statistical Science)

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

 

 

I took Prof. Faraway's course 10 years ago in which we used a very early version of this book. I am glad to see that this book is finally in print. For years I have been using Prof. Faraway's notes teaching a graduate course on regression. What I like this book most is its emphasis on practical usage and pitfalls of linear models. This is how linear regression and statistical modeling should be taught but I am not aware of any other textbooks doing the same thing.

This book, just as its author, is thin, clean and concise. It does not always explain and reveal things in full detail. If you are smart, like to think and explore data using R/Splus, or have a good mentor, this book fits you well. Its discussion on principle component analysis, in my opinion, is a little bit weak.

This volume is the best hands-on R book I could find which opens the door to lm() in R. The book is thin and the contents somewhat dense - there is no room for hand-holding: you need to learn the basics of R and statistical modeling elsewhere. But if you meet the prerequisites, buy the book, read it, and most importantly, TRY THE EXERCISES!

GOOD POINTS
- exercises (deserves 5+ stars for learning concepts with real data)
- short chapters, so you can quickly test your understanding via exercise
- chock full of R examples that you can try with library( faraway )

BAD POINTS
- proofs are not rigorous enough for mathematicians, but too dense for practitioners (who would prefer more intuition)
- helps if you have played with R a bit and understand basic statistics
- no answers to exercises

The two things you typically want in a book, this one has -- it is clear and concise. I'm a stickler for how things are explained and this book surpassed all expectations, explaining topics elegantly. Not only are methods explained well, but so is how to interpret data as well as general advice and guidelines on fitting models and checking that assumptions are met. When reading this book I feel like I am getting a lot more than just how to fit linear models but how to analyze and judge if a model is appropriate, which is a crucial step to fitting.

I've read a good portion of the book, reading the first several chapters and skipping around more on a need-to-know level for the other topics. Below is a list of the chapters:
1. Introduction.
2. Estimation
3. Inference
4. Diagnostics
5. Problems with Predictors
6. Problems with the Error
7. Transformation
8. Variable Selection
9. Shrinkage Methods
10. Statistical Strategy and Model Uncertainty
11. Insurance Redlining -- A Complete Example.
12. Missing Data
13. Analysis of Covariance
14. One-Way Analysis of Variance
15. Factorial Designs
16. Block Designs

I will inevitably be buying what is like the second volume of this book, "Extending the Linear Model with R" as needs arise.

Very clear, very concise. The author moves from important point to important point in a smooth manner. Though, there is definitely room for some more detail, that's not what I bought it for. I wanted a good summary of linear models and how to use R to fit and analyze them. That's exactly what I got. If I want more detail, the author places references to appropriate books throughout the book.

I got this book because I needed to use robust techniques in regression and ANOVA. I could not find much in the R-documentation to help with this.

The book is excellent in the sense that it guides the reader through a number of fairly useful techniques using linear models and generalized linear models. I adopted the strategy of starting with the first chapter and working each example in R as the author presents it. I've learned a lot about how to use R doing this. It took me a while to get used to the "unified" linear model idea- viz., regression and ANOVA are based on the same linear model. The book has a number of excellent examples that demonstrate the power of R and show the reader how to exploit various library functions.

Overall I like the book. It is an excellent introduction to how to use R in one's linear model analyses. It is long on "here, type this in and you'll get....." and short on the theory/principles behind each technique.

One aspect of R I have not yet cracked- how do I (easily) specify a linear model that includes only interactions up to level 2 or 3? I currently enter the formula as Y~A*B+A*C + ... and so on. Must be an easier way.

The discussions on robust methods are good, but (understandably) short. Use of the trimmed least squares and quantile regression techniques as well as a discussion of using bootstrap methods to get a confidence interval are, again presented as "here, type this in and..." At least this book got me headed in the right direction. There is also a 2- or 3-page summary of basic R commands to help the neophyte get started with R.

A good book, but could be better with some presentation of the theory behind each of the techniques presented.

very bad book in general, it skips a lot of the steps in R.
This book assumes you're pretty proficient in R, there are no solutions to the problems anywhere so you can't check your work)

I have not read this version of the textbook yet, but I took his class. The textbook he used for that was his working draft of this book. It is very practical, using the free R software (based on S-Plus) to learn and amplify the concepts with example data sets. This allows deap learning of the methods and easy application to one's own uses. R is not that easy to use, but working through his book, one can learn all the parts of it needed.

The best thing in his approach is the coverage of model uncertainty. Is the model applicable or is it not? At least constantly thinking of this is critical, so his constant reminder of that is more realistic than most approaches.

The book is accessable to anyone with basic statistics background, but a bit of mathematical statistics would really help for a deep understanding.

This book is a concise introduction to linear models which presents a cookbook approach to data analysis and modeling. I find it very surprising that this book has so many good reviews. At first glance it looked like a very good book, but after working with it for several weeks I began to find it boring, difficult to follow, and not comprehensive enough--to the point of being outright aggravating. The thoughtful reader will undoubtedly have many questions that this book does not adequately address. This book could easily turn a reader with less initiative into a poor statistician.

I would greatly appreciate a more critical approach to the subject. Unfortunately, statistics is a field in which many things are done a certain way just because they have been done that way for many years, or because that way is particularly convenient. Like much of the field of statistics, this book is built on shaky foundations. This book does not do a good job of justifying the particular approaches used, nor does it really explore many alternatives or different ways of looking at things.

Also, it's not clear what the intended audience of this book is. The "cookbook" approach seems to be appropriate only towards people in other subjects with minimal math background and little desire to do research in the topic. But at the same time the author seems intent on explaining the underlying ideas and mathematics behind each technique. While I appreciate this approach, the explanations do not satisfy me. Also, while the authors do seem to make considerable effort to make the mathematics intuitive and accessible, I still found this book to contain many passages where the use of equations or mathematics is confusing or unclear. For example, I found the statement and proof of the Gauss Markov theorem to be particularly unnatural and unintuitive. The book does not contain many mathematical exercises which can help the reader become adept at the particular notation and perspective used.

The use of datasets and exercises related to these datasets I find much more useful. I just wish that the book would do a better job of showing how theory can interplay with practical considerations and how abstractions ultimately arise from structure and organization that exists in real world data.

Bottom line? I would recommend against using this book for any purpose. It was used as the textbook for a course I took and I did not find it at all useful. I think that this book could be improved either by cutting out some of the theoretical discussions (which are confusing and yet unsatisfying) and shifting the focus of the book more towards data analysis and modeling, or...on the other hand, shifting away from this and including deeper discussion of theory. In either case, the book needs more skepticism and needs to abandon the cookbook approach if it is going to be anything other than misleading. As a better book, with both more explanation, and with a more critical approach, I recommend "Introduction to Regression Modeling" by Abraham and Ledolter. At a more advanced level, Freedman's "Statistical Models: Theory and Practice" is another good alternative.

Illustrates how to do statistics with R. Book can be used with students or as a learning tool for those in statistical research areas who want to see what R can do.

Covers a lot on the new Regression Analysis in a very practical way and easy use of R.