Customer Reviews

Bayesian Computation with R (Use R!)

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

 

 

The good: The first three chapters gives the reader a nice introduction to using R for Bayesian statistics and some well worked out examples: a necessity when dealing with a program that one is unfamiliar with. The text does a decent job of complementing the material found in another text on basic Bayesian methodology such as Gelman et al. (2004) or Carlin and Lewis (2008). Furthermore, Jim Albert is a great writer and presents the material well.

The Facts: Towards the latter half of the text the author begins to use a program from the 'Learn Bayes' package entitled "Laplace". It is of my belief that this black box could be elaborated on some. I had some trouble getting many of the examples from the text as well as exercises from the sections to run simply because of this black box. None of nine graduate students working together and independently were able to get this function to perform its duties on a regular basis. However, the examples and problems were instructive.

The Opinion: I was not a fan of the functions from the Learn Bayes package and did not feel as though the reader gained an adequate background on how to program R to perform Bayesian methods on his/her own. The book, I believe, relied to much (in the latter half of the text) on the functions of the Learn Bayes package.

Overall the text is great resource to complement another text. The only real `issue' I had with this text was not the text itself but rather the "Learn Bayes" package. If you are looking for a resource for R this might not be the right book. As a quick and dirty introduction to Bayesian methods using R (as the title suggests) this isn't a BAD text.

Jim Albert is a great teacher and an excellent writer. The R language is becoming one of the most used languages by statistical researchers. This is because it has many similarities to S and can be used freely, Jim makes R easy to learn for statisticians in this book. One of the big breakthroughs in Bayesian statistics over the past 2 decades was the implementation of complicated priors and hierarchical models through the Markov Chain Monte Carlo (MCMC) algorithms. The leaders is this filed created free software called BUGS (for Bayesian Analysis Using Gibbs Sampling). Gibbs sampling is one of the most commonly used MCMC algorithms. Statisticians using this software have been able to provide more satisfactory solutions to many basic and complex problems using these tools. After Windows became the dominant operating system on personal computers WINBUGS was born. This is a version of BUGS that uses Windows as the operating system and takes advantage of Windows many nice features. Now for the first time to my knowledge Jim Albert show the reader how to incorporate the BUGS technology in the framework of R programming. This can only add to the practical use of Bayesian methods among statisticians for research that advances both the theory and applications. In the late 1990s I was working in the medical device industry where a number of clinical trials were being analyzed using the MCMC methods. Jim deserves a great deal of credit for moving Bayesian statistics into the framework of R!

This is a great book that introduces practical Bayesian computing for scientists and quantitatively oriented people. Good sections on MCMC and other aspects without getting too mathematical (as opposed to being statistical - Does not mean that you won't find any symbols). Having said that, please be at the level of Casella/Berger on the (frequentist) mathematical statistics level and one of the following books should serve as a good companion for Bayes theory - Peter Lee (2004 - Great introduction), GCSR (Gelman et. al.) or Carlin and Louis. If you want to learn further details of the computational algorithms, MCSM by Casella and Robert is an excellent reference.

The book starts out by introducing us to R and then the Bayesian way of thinking and analyzing data. Up until chapter 5, we learn how to summarize posteriors when functional forms exist and how the various author-created functions serve the purpose. (The author's LearnBayes package contains many excellent functions that can be used in a wide variety of situations). Chapters 6, 10 and 11 form the core of how you perform MCMC and the various algorithms behind it. BRugs is introduced as well. I would also recommend the author's website and his excellent blog for learning 'Bayes'. http://bayes.bgsu.edu/bcwr/

Good resource if you are motivated enough, but you definitely need a companion book on Bayesian Statistics if you are not already well-versed in the theoretical aspects of these techniques. Great book for the Bayes-curious statistico... Of course, if you are reading this review you don't have to be told how great R is. Price has dropped 20% since it first came on the market. I'd say, a steal at 40 bucks.

Great book. If you work through the examples, this book will move you to very near the top of the R learning curve and, more importantly, race you to the peak of the Bayesian curve.

First two chapters are easy to follow, after that its down-hill. A lot of equations are stated without an explanation, no clear expositions. Writer falls into the trap of writing a book for himself and not the student. Nothing is explained clearly. No one in the 'real' world uses R, only mathematics departments, so you want be applying these functions in work related environments. No answers given to chapter questions. The book is called 'Bayesian computation in R', only if you understand Bayesian computation very well and need to see some examples in R. Otherwise your not going to learn Bayesian stats with this book.

I am giving only one star. I would like to give this book a zero star.

The author uses his own built-in data set and his own program package "LearnBayes" in this book. No clues given on how he got this data into R.

The main problem with R is getting the data into R.

Same thing can be said about Casella's books-- utterly useless.

All these books are overpriced. Being professors, they can force the students to buy these books.

Albert could have posted his version of the book on his web site as a pdf document and let the readers download it for $1 or $2.

A somewhat useful book is the one by Michael Crawley ("The R Book").

What R needs is a simple command to get the data into R.

R may be free. But not much of practical use for real problems unless you already are a programmer or a PhD Mathematician.