Customer Reviews

The R Book

5 star:		(16)
4 star:		(7)
3 star:		(4)
2 star:		(3)
1 star:		(3)

 

 

This book is both ponderous and expensive, so my decision to buy it was predicated on the dual claim that it's 'the first comprehensive reference manual for the R language' and `ideal for novice and accomplished user alike'. As an R beginner and non-statistician (with some long-ago training therein) pressed into scientific data analysis on a regular basis, I wanted a comprehensive reference that covers both the R language and theory behind modern applied statistical methods.This is no small undertaking, but Crawley succeeds reasonably well at the task.

The book contains 27 chapters. The first 5 chapters cover subjects like getting started, essentials of the R language, data input, data frames, and graphics. A lot of the information in these chapters is freely available online at CRAN, or may be queried from within R itself. Still, I find it useful to have this info as part of any desktop reference, and most books on R are similarly equipped. I found nothing lacking here.

Chapters 6-8 cover tables, mathematics, and classical tests. In the mathematics chapter, you'll be introduced to a wealth of math and probability functions, as well as the basics of matrix algebra. If your statistical training centered mainly on the basic normal, student's t, Fisher's F, poisson, and chi-square distributions, get ready for an education. The author's presentation of this material is both in-depth and well articulated.
Chapters 9-20 cover statistical modeling, regression, ANOVA, ANCOVA, GLM, count data, count data in tables, proportion data, binary response variables, GAMs, non-linear models, and mixed effects models.Chapters 21-26 address more advanced topics of tree models, time series analysis, spatial statistics, multivariate statistics, survival analysis and simulation. The author's discussion of statistical models, ANOVA, GLM, and mixed effects models (the four chapters I have dug into thus far) covers theory as well as practical application inside R. Chapters are supplemented with worked examples drawn from various R data libraries. The R code used to generate solutions is presented as well, although I found it difficult to integrate because Crawley is using the R console interactively and snippets of code are spread out over many pages. Yes, you can download a data library, type in the code presented in the book, and get the same output. The difficulty arises in making the transition from textbook example to efficient and statistically valid processing of real- world data. If you're new to object oriented programming, this book will not teach you how to program in R. Only practice and good example can do that. I still struggle with some R programming basics and this book did not help at all.

Oddly, the book ends with a final chapter 'Changing the Look of Graphics'. Seems like this should be part of chapter 5 'Graphics'; it's a mystery why this was broken out as a separate chapter and stuck at the end.

The book contains numerous typos that suggest a lack of proofreading. Also annoying is the author's predilection for cross-referencing, such that one is constantly being advised to 'refer to page ...' for more info. Furthermore, the author profanely suggests Word as a text editor (yikes!). There are excellent text editors freely available for R, but Word isn't one of them. I use TINN-R, but there are other options. Also, options for managing R output are given short shrift. I use Notepad++, a tabbed, free text editor which is similar to TINN-R, but external to R. FYI, Notepad++ will also read SAS output in its native format, so one can easily review, compare, and extract information without invoking an R or SAS session.

Be advised, this book has created some controversy within the elite, tight-knit R Core Development group. The book was reviewed in the October 2007 issue of R News, available online (thumbs down). Crawley evidently is not part of the R Core Development 'inner sanctum', so the book's rather grandiose claim as 'the first comprehensive R reference manual' has engendered some criticism from that group. Other criticism about R expressions, the author's advice regarding use of certain R functions, and use of specific R packages may be found therein. Read the review then make your own judgment. As it stands, I don't consider this book to be an authoritative reference on either statistics or the R language, but it does offer an inclusive survey of both. If you already own a good statistics text, are familiar with object oriented programming, and only need a reference explaining how to get started programming in R, you'll save money by buying An Introduction to R by Venables and Smith. Amazon's wallet- friendly price: $13.57. Or you may download a free PDF version from the CRAN website.

I'll give the book four stars. It has some flaws (a second edition would be welcome), but overall constitutes a useful addition to the R literature. As for programming, I'm eagerly awaiting Braun and Murdoch's 'A First Course in Statistical Programming in R'. There are enough books on R-based statistical analysis in the vein of Crawley and others; we need a book that teaches programming and the latter should fill the gap nicely.

Given the length of this book, and the list of contents covered, I had the highest expectations about it.

After spending 2 intensive months reading it, I have mixed feelings. Positive points are the large number of statistical models and methods described. The R examples are useful to follow the explanations, and the writing style is comprehensive. I agree with some reviewers in that the linear models section (Chaps. 9-19) is the most useful one. The last Chapter also presents useful tricks for dealing with graphs in R.

Unfortunately, I have 2 important complaints. The first one is about the presentation of contents: simply CHAOTIC. The author systematically abuses of cross-references. You will find sentences like "here we present an example of [method XX] that will be introduced on page XXX" throughout the entire book. This is disappointing, since it forces the reader to constantly move back and forth, looking for the relevant info. There is no point in presenting an example based on a method that you haven't introduced yet. Examples should be autonomous, and not frequently taken from previous data sets "already used in page YYY".

The second complaint derives from the previous one. The book is hard to use as both a reference manual and a companion for undergraduate or graduate students. Disregarding the comments from the author, if you don't have a solid theoretical background in statistical inference, regression analysis and linear models, you won't get very much benefit of this book. The author completely lacks of a rigorous, structured method for presenting new concepts. Even worse, important definitions and concepts are usually hidden in between of examples that has nothing to do with them.

In summary, if you already have a good theoretical background in statistics, this could be a useful add-on to your bookshelf (though be ready to spend a lot of side tags to map important concepts for later).

If you're looking for a introductory book with R, Springer has just published a second, expanded edition of the classic book by Dalgaard. If you're looking for a definitive reference manual of statistical methods illustrated with R, you will have to wait for something else, or look for specific titles (Like Faraway's "Linear Models with R"). For Ph.D. students looking for a comprehensive an up-to-date book on statistics with R, to improve their skills quickly, I still recommend the second edition of "Data Analysis and Graphics Using R", by Maindonald and Brown.

I have been using R for about 5 years and for the last 2 years have I been using it regularly. This book might have helped at the early stages of learning R, but at the moment I have learned not to trust this text. I am now seeing questions on the r-help mailing list related to the disorganization that others have commented upon.

The problems start early. For Windows users he uses doubled backslashes without explaining why these are needed or the alternative. He gets the distinction between the two indexing operators wrong when he states that one does not use "[" with lists (and fails to note that dataframes are in point of fact lists.). After starting with the R concept of vectors, the author introduces functions which have list arguments before even describing either lists or the list() function. I could go on and on. Going much further on to the section on count data he compares a dataset to the theoretic Poisson distribution noting that the counts at the low and the high ranges are larger than expected, calling this "highly aggregated", whereas most statisticians would call this over-dispersed. In the next paragraph he makes the opposite statement but then says that the overly dispersed data "shows randomness". I was further bother by his frequent use of the dangerous practice of using attach and the confusing practice of naming objects using names that were also function names such as "exp" and "data". I get the sense that the R News reviewer gave up noting errors and decided instead to move quickly on to recommending "Modern Applied Statistics" by Venables and Ripley. I would also suggest Harrell's "Regression Modeling Strategies".

People sometimes complain that the R documentation in the help pages is difficult, but in my experience the R help pages are far more precise than Crawley's versions of R. I would avoid this book due to a) its disorganization, b) its misleading advice regarding R syntax, and c) its statistical errors. It probably has some value for its large number of worked examples but there are much better books available.

Flatly, this tome is poorly written, poorly organized, and has clumsy, sometimes erroneous coding examples. It attempts to cover a wide array of applications, but I cannot guess the intended audience. I can imagine many people who could use a book on R, but "The R Book" would fail all of them:

1. Scientists who want to do some statistics and need an introduction to programming. The examples are too opaque, and the writing too dense (beyond simply being bad) for this. It is not at all a statistics book, either, so is unhelpful for learning that. Colleagues of mine who attempted to learn programming using this book were all very frustrated and quickly stopped consulting it.

2. Statisticians who need an introduction to programming. The book does assume some knowledge of stats, but deals with fairly unsophisticated problems, and again, is just not written for beginner programmers.

3. Competent or semi-competent programmers who are learning R. This is the group I fell into, and the book offers nothing the internet doesn't. It doesn't go into any depth on the structure of the R language and its coding examples are awful: They're poorly formatted (a non-monospace font with inadequate spacing) and sloppy (several examples introduced variables that were never used). Once I realized the apparent inaptness of the authors in programming, I ceased to trust them on anything.

It's probably really just 4 stars, but compared to other R books I've seen it's 5 stars. It's comprehensive and relatively easy to follow. It covers a lot of topics. The code is easy to follow. The index could be better, but it's not bad.

It is an excellent book if you want something both to bring you up to speed, and then to serve as a comprehensive reference.

A good approach to collecting R books would be to start with this book, and then if you outgrow it in certain areas, obtain topic-specific R books in such areas modeling, data manipulation, or graphics as supplements.

I'll make this short and sweet: On the one hand, "The R Book" is a hugely encyclopedic reference to the R language. If there's something in the basic R language, it's probably in this book. On the other hand, this book is horribly organized and horribly written. It needs an editing job of epic proportions. Its disorganization and bad writing make it particularly poorly suited to the task of learning R. If you are an R user, this book probably belongs on your bookshelf somewhere. But make some room on that shelf for a few other, more helpful books, because this one is just not the place to learn R.

The biggest advantage of this book is its sheer size and comprehensiveness. If you were only going to own one book on R, this would probably be a good choice.

But the book is flawed. First, no book, even one this size, can cover all the material it purports to cover. Second, the material on what might be called the 'base' of R is poorly organized, and the index could be better. If you want to know how to do something in R, it's probably in here. But it may take you a while to find it. Third, the material on programming is quite limited.

Still, I would recommend that R users, particularly those relatively new to the language, get this book. I use it often, and usually find it useful.

This is a great book for someone new to R, who also has some background in mathematics and programming. But, it is also helpful for the complete beginner because it shows you, step-by-step, how to load data into the R environment so you can actually get started using R, even if you don't have a nearby R mentor to help you out.

The writing is clear and accessible with examples provided for nearly all of the R software tools discussed. Also useful is that the author not only tells you which tools to use, but he also often says why they are important.

It's a thick book, but if you take the time to work your way through it, you should actually be able to use R to solve real world problems without external guidance from a R veteran. Check it out!

This review is about issues specific to the Kindle edition.

This book contains many internal cross-references ("see page 234") which are useless in the Kindle edition -- they are neither hyperlinked nor is there any way to jump to a numbered page (in fact, Kindle does not support page numbers at all). This is also true of the index. It is of course possible to use free-text search, but that doesn't help when you want to find the defining or most important references to a particular term.

The table of contents *is* correctly hyperlinked, but only at the chapter level. Given that most chapters are over 20 pages long, and several are over 60 pages long, that is not enough detail -- an analytical table of contents would have been much more helpful.

There are also some minor formatting problems, but they are negligeable compared to the above issues.

It's about time that software like R became available and popular, it is clearly the way to go. I've been learning it on my own for a few months now with the help of about a dozen online .pdf file manuals, and sometimes using the Nabble forum for R. I finally broke down and bought this book, and wish I would've bought it in the beginning. If I were to write a book on this subject, this is pretty much exactly what I would do...except I'd like to see a companion volume that explores the numerous packages, maybe with an emphasis on Bayesian methods, such as in the packages arm, boa, coda, MCMCpack, MNP, R2WinBUGS, etc., but hey, that's me.

If you've had it with other software that doesn't let you do everything you'd like to do, then I highly recommend R, and The R Book for starters.