Markov Chain Monte Carlo in Practice (Chapman & Hall/CRC Interdisciplinary Statistics)

Great product, fast delivery, thanks!

This book is barely readable.

Gilks, Richardson and Spiegelhalter edited this marvelous collection of papers on applications of Markov Chain Monte Carlo methods. There has been a big payoff for Bayesians as this method has been a breakthrough for dealing with flexible prior distributions. Most (but not all) of the articles deal with Bayesian applications. The editors themselves start out with an introductory chapter that covers the basic ideas and sets the stage for the articles to come. They provide many references including several of the articles in this volume.The list of authors is quite impressive and many interesting examples are presented. The editors themselves contribute to other chapters. Spiegelhalter and Gilks co-authored a chapter on a Hepatitis B case study with Best and Inskip. Gilks has a chapter on full conditional distributions and co-authors a chapter on strategies for improving the MCMC algorithms. Richardson contributes a chapter on measurement error.George and McCulloch deal with the use of Gibbs sampling to choose variables in a model based on a Bayesian approach. Raftery also has a chapter on Bayesian approaches in hypothesis testing and model selection. Green covers image analysis. There are many others (25 chapters in all). This is a great reference for anyone interested in MCMC methods.The BUGS (Bayesian inference Using Gibbs Sampling)software was developed by Spiegelhalter, Thomas, Best and Gilks to implement Gibbs sampling in a variety of contexts. They illustrate its use along with the diagnostic software CODA in the application in Chapter 2. It is also mentioned in various other chapters in the book. There is currently a version called winBUGS which is designed for Windows operating systems.Before jumping into the use of MCMC a user would be well advised to study this book.

First, I'll like to comment on the termiology. I'm PhD specializing in stochastic simulation in operations researcn and I've found the book is written in a language that's not quite standard (it might have something to do with his background in Statistics). Some people may argue that "names" are just "names" but it could cause confusion. And, in the chapter of stochastic approximation, the author failed to mention a couple of well-known existing methodology (somehow show a poor literature review in the field.) Strong emphasis has been given on importance sampling on that particular chapter, but author failed to mention in what context will importance sampling work. If you assume Bayesian approach and have prior on the parameters, then it works. But, if you're a frequentist, it's not necessarily working for your model.Going back to the first chapter, I found the construction of MCMC is presented much more clearly in Sheldon Ross's Probability Model rather than this book.

We recommend this book to anyone who is interested in learning MCMC methods. Contains a excellent selection of practical examples. Christopher Gordon and Steve Hirschowitz