Product Description

This Bayesian modeling book is intended for practitioners and applied statisticians looking for a self-contained entry to computational Bayesian statistics. Focusing on standard statistical models and backed up by discussed real datasets available from the book website, it provides an operational methodology for conducting Bayesian inference, rather than focusing on its theoretical justifications. Special attention is paid to the derivation of prior distributions in each case and specific reference solutions are given for each of the models. Similarly, computational details are worked out to lead the reader towards an effective programming of the methods given in the book. While R programs are provided on the book website and R hints are given in the computational sections of the book, The Bayesian Core requires no knowledge of the R language and it can be read and used with any other programming language.

The Bayesian Core can be used as a textbook at both undergraduate and graduate levels, as exemplified by courses given at Université Paris Dauphine (France), University of Canterbury (New Zealand), and University of British Columbia (Canada). It serves as a unique textbook for a service course for scientists aiming at analyzing data the Bayesian way as well as an introductory course on Bayesian statistics. The prerequisites for the book are a basic knowledge of probability theory and of statistics. Methodological and data-based exercises are included within the main text and students are expected to solve them as they read the book. Those exercises can obviously serve as assignments, as was done in the above courses. Datasets, R codes and course slides all are available on the book website.

Product Details

• Amazon Sales Rank: #649808 in Books
• Published on: 2007-02-02
• Original language: English
• Number of items: 1
• Binding: Hardcover
• 258 pages

Review

From the reviews:

"The matching of each computational technique to a real data set allows readers to fully appreciate the Bayesian analysis process, from model formation to prior selection and practical implementation." (Lawrence Joseph from Biometrics, Issue 63, September 2007)

"Recent times have seen several new books introducing Bayesian computing. This book is an introduction on a higher level. ‘The purpose of this book is to provide a self-contained entry to practical & computational Bayesian Statistics using generic examples from the most common models.’ … Many researchers and Ph.D. students will find the R-programs in the book a nice start for their own problems and an innovative source for further developments." (Wolfgang Polasek, Statistical Papers, Vol. 49, 2008)

"This text intentionally focuses on a few fundamental Bayesian statistical models and key computational tools. … Bayesian Core is more than a textbook: it is an entire course carefully crafted with the student in mind. … As an instructor of Bayesian statistics courses, I was pleased to discover this ready- and well-made, self-contained introductory course for (primarily) graduate students in statistics and other quantitative disciplines. I am seriously considering Bayesian Core for my next course in Bayesian statistics." (Jarrett J. Barber, Journal of the American Statistical Association, Vol. 103 (481), 2008)

"The book aims to be a self-contained entry to Bayesian computational statistics for practitioners as well as students at both the graduate and undergraduate level, and has been test-driven in a number of courses given by the authors. … Two particularly attractive aspects of the book are its concise and clear writing style, which is really enjoyable, and its focus on the development of an intuitive feel for the material: the numerous insightful remarks should make the book a real treat … ." (Pieter Bastiaan Ober, Journal of Applied Statistics, Vol. 35 (1), 2008)

"The book is a good, compact and self-contained introduction to the applications of Bayesian statistics and to the use of R to implement the procedures. … a reader with a previous formal course in statistics will enjoy reading this book. … the authors are not shy of presenting such complex models as hidden Markov models and Markov random fields in a simple and direct way. This adds an edge to a compact and useful text." (Mauro Gasparini, Zentralblatt MATH, Vol. 1137 (15), 2008)

"This book’s title captures its focus. It is a textbook covering the core statistical models from both a Bayesian viewpoint and a computational viewpoint. … There is a discussion of choice of priors, along with math to derive the priors. … The book is being actively used as a textbook by a number of university courses. … The course level is graduate or advanced undergraduate. Solutions to the exercises are available to course instructors … . In conclusion, the book does what it does, well." (Rohan Baxter, ACM Computing Reviews, December, 2008)

Customer Reviews

Book has a purpose
I have used 'Bayesian Core' to teach Bayesian statistics to a class of Masters students majoring in finance, statistics or business with an undergraduate mathematics/statistics background and found it to be quite a good book for this purpose. I would have to disagree though with the previous reviewer (A.L.H Mayne) and suggest that this book could also be good for a practitioner, provided they have some prior statistical and mathematical understanding. A limitation for self guided study is the absence of solutions to exercises or hints necessary to ensure understanding for some of the exercises. I would also have to disagree with the previous reviewer over what is or what is not discussed in the book! The statement "conjugacy is mentioned in exercise 2.10 on page 22 with no discussion" is simply not true, the entire page preceding this exercise discusses conjugate priors in particular and they are subsequently used and outlined in the following chapters. There is also considerable attention paid in the book to the use of improper priors, in particular with respect to the implications of using improper priors for the estimation of Bayes factors (Chapter 3). A more thorough description of Jeffrey's Lindley paradox could be provided but a more complete discussion of this seems outside the scope of the book. Similarly, outlining the "historical antecedents ..." and more "...theory" about Jeffrey's prior than what is already provided, while interesting and important in its own right, is not necessarily mandatory reading for the student or practitioner who seeks a practical introduction to the Bayesian approach. To highlight "The implementation of the Monte Carlo method is straightforward" without adding the next few words used in the book "at least on a formal basis" seems terribly insincere and pedantic.
What the book is? The book presents a Bayesian approach to the analysis of topics commonly analysed in statistics designed to allow the reader to quickly grasp the essential elements of Bayesian principles and to put this into practice with examples using R code (simple computing syntax) provided on the website. There are surely limitations of this approach (albeit also acknowledged by the authors!) namely a less than full treatment of topics and of theoretical derivations for approaches. Some of the exercises are also difficult and these exercises could well do with hints being provided or a short statistical annotation.

a versatile book
"Bayesian Core" attempts to balance Bayesian theory, computations and applications in a compact book of about 250 pages intended for teaching and learning modern Bayesian statistics. It succeeds in its ambitious goals to some extent but does have some shortcomings. The book has some nice features that I like:
(a) Exercises are placed at strategic locations within the main body of a chapter rather than put collectively at the end. This helps the student to consolidate key ideas before moving on.
(b) In each chapter, one or two real-world data sets are adopted as a common thread linking ideas and techniques within the chapter. This helps to demonstrate application in a timely and tangible manner. Computer implementation is also aided by the provision of pseudocode or R code.
(c) Warnings about common pitfalls are provided and highlighted.

Generally, I find "Bayesian Core" to be a highly versatile book that is useful at different levels. At the level of introductory-to-intermediate Bayesian statistics, I taught a course in 2006 and 2008 for third and fourth year undergraduate students (with second year statistics background) based on the book. The course comprised 24 lecture hours plus 12 supervised computer lab hours and I was able to cover most of the material in chapters 1 to 4. As an instructor, I had access to slides and solutions that accompanied the book and found these to be extremely helpful for conducting the course. I would not recommend the book for self-study to someone who is encountering Bayesian statistics for the first time (self-study from the book is quite feasible for someone who already has some background in Bayesian statistics). However, when used in an instructed course, the book and its accompanying resources worked quite well. The response of the students to the course was quite positive, with the majority finding it refreshing, interesting and challenging. Some students found the book to be a useful resource both during and beyond the course. Because the book relies on one or two data sets to develop and illustrate ideas within a chapter, examples are mostly confined to those key data sets. To give students a broader perspective, I chose to supplement the course with examples and R exercises taken from Jim Albert's book, "Bayesian computation with R".

At the advanced undergraduate level, I have used material in the book for final-year projects. Much of the book's contents can easily be adapted and packaged into interesting projects. At the graduate and research levels, the book is a handy resource that provides concise descriptions of key ideas and algorithms, including some important ones that are not commonly found elsewhere (an example is pivotal reordering for mixture models).

My wish-list of improvements are:
(a) Some of the exercises can be difficult and can do with more hints.
(b) Some of the contents of the slides can be incorporated into the book to improve clarity.
(c) Additional exercises, preferably involving data sets other than the key ones used in the chapter, can be provided at the end of each chapter so as to give a more diverse exposure to applications and examples.

a different approach to teaching Bayesian methods
First, it is only fair that I state upfront that the publisher Springer-Verlag provided me with a copy of the book and asked me to give a candid account of the book. Obviously the publisher was hoping for a favorable review to balance the negative review by Mr Mayne. Mayne's review stood alone on amazon at the time. The editor at Springer knows me personally and knows that I am a well established statistician who has done many capable reviews on amazon and in the journals. Springer-Verlag also at one time offered me a contract to write my bootstrap book with them.

In the email message, the acquisitions editor suggested that Mr Mayne provided a review that had incorrect statements and at the time stood as the only customer review of the book. After reading the review I recognized that Mayne had a good knowledge of Bayesian statistics and might have a few valid points. While I had not yet read the book I felt that Mayne was familiar with many excellent texts on Bayesian statistics and I agree that Jeff Gill's text is one of the best introductory texts on Bayesian statistics for non-statisticians. It covers the MCMC method about as well as any text I have seen.

On the other side I took a short course on Monte Carlo methods from George Casella and Jeff Gill using the book written by Casella and Christian Robert, one of the authors of Bayesian Core.

Now there is a second review which refutes some of what Mayne says that was unfair to Marin and Robert. So the need for my review may be less important. However, now that I have reviewed the book I think I have comments that will show that I have a very favorable impression of the text but also understand the points of view of both previous reviewers.

I believe that Mayne was bothered by the claim of the authors that the text is self-contained and suitable for undergraduates and graduate students and practitioner's who are not statisticians. The text is shorter than most others and the term core in the title suggest that it contians the essentials but is not all inclusive. I have some sympathy for Mayne's position. I do feel that the book has the elegance and terseness that is common to the French writing style. Self-contained means that it includes all the fundamentals needed to obtain the results in the text. I do think the book is self-contained but that does not make it elementary or easy to follow. As with the other books that Robert has authored or coauthored the text is very well-written. I do think that it is a little advanced for undergraduates and is best suited to be a graduate text in applied Bayesian analysis for statisticians or statistics majors. It could be used by practitioners for self-study but with difficulty.

There are some special features of the book that I like and feel should be pointed out. R programming is introduced as a way to do Bayesian calculations and provides a reasonable alternative to Winbugs though both approaches involve software that is essentialy free. Jim Albert has written a text that does a better job of demonstrating the virtues of the use of R and the CRAN library of R routines for Bayesian analysis.

Each chapter contains examples and exercises and the exercises are introduced throughout the chapter rather than at the end of the chapter. This helps the reader recognize the need to not rush through the chapter by pausing to try the exercises and see how well the material is understood.

There is a great deal of emphasis on computational methods and although Jeff Gill's text may give a better and more detailed account of MCMC in layman's terms, Marin and Robert do a fine job of covering the topic and many special applications. Capture - Recapture experiments using Bayesian methodology are presented in the text and I have not seen the topic covered in other texts. The authors also do a great job of presenting mixture models from the Bayesian perspective as well as time series and generalized linear models. Probit modeling is also discussed. Another topic important in modeling is subset selection and the authors illustrate this topic with real Bayesian examples.

I also feel that improper and uninformed priors are dealt with in depth in the text. The mathematics is important to help the readers understanding of the theory, methods and results but also makes it a little more difficult for some practitioners and some undergraduates.

MCMC methods were rediscovered by Geman and Geman to solve problems in Bayesian imaging analysis. Image segmentation, image reconstruction and improving resolution are all important topics that are naturally approached by the authors.

All in all I find this to be a very useful text that is well-suited for a graduate level course covering important topics without being overly elaborate.