Markov Chain Monte Carlo (Chapman & Hall/CRC Texts in Statistical Science)

This is an excellent book. It describes the material clearly, using plenty of cogent examples. I gladly recccomend this book.

I have several negative comments before a full review and these apply to the first edition of their text. First, most of Chapter 4 (sections 4.1 - 4.5) is lifted from Chapters 1 and 2 of Hoel, Port, and Stone (1972) and reorganized without any attribution - even the notation and examples are the same. I find this offensive. Second, they could really use an English-speaking editor to clean up the way the text reads, I can't name the number of times I saw "highly dimensional distributions" and other annoying English. The editor of this book should be fired, if there was one. Of course, in the Preface it does state "The book grew out of lecture notes in Portuguese prepared for a short course on the topic taught at the XII Meeting of Brazilian Statisticians and Probabilists, held in Caxambu (MG) in August 1996." Consider yourself warned.That said, the book is informative, with the following major outline:Chapter 1: Stochastic SimulationChapter 2: Bayesian InferenceChapter 3: Approximate Methods of InferenceChapter 4: Markov ChainsChapter 5: Gibbs SamplingChapter 6: Metropolis-Hastings AlgorithmsChapter 7: Further Topics in MCMCThe first three chapters are informative, but you won't learn the theory for the first time from this text. Chapter 4 focuses primarily on discrete parameter, discrete sample space Markov processes - Markov chains. It's not as clear as the original Hoel, Port, and Stone from which it has been lifted and reorganized. It is useful as review if you already have a solid grasp of Markov chain theory. At the end of the chapter there's some basic prose about extending the major results of Markov chain theory to general state space, discrete parameter space Markov chains so that continuous limiting distributions can be simulated. Chapters 5 and 6 are adequate for understanding Gibbs sampling and the Metropolis-Hastings algorithm for constructing Markov chains that have a specified limiting distribution. The book is adequate for its stated purpose, though I found no explicit coupling of the MCMC theory to Bayesian inference problems in the text...notwithstanding the title of Chapter 2 (see above) and the subtitle of the textbook.

This book is very self-contained and provides intuitive explanations and illuminating examples. Very good for self-study.

If you are a Bayesian statistician or maybe if you want to understand this statistical paradigm, this book is not only necessary but indispensable. Professor Gamerman and Lopes present an excellent study about the simulation techniques that we need to implement if we want to draw samples from posterior distribution when we don't have this in a closed way. This book brilliantly analyzes the most basic MCMC methods like Gibbs Sampler and Metropolis Hastings, on the other hand the reader shall find a lot of funny examples and some applications.

I am a fifth year graduate student in Psychology. I was assigned this book as part of my readings for candidacy. I also had this book for a seminar course on Bayesian methods. The first time I encountered this book during the course, I must admit that I found it difficult to read, so I ended up not being motivated enough to go through with all the assigned readings. I have taken 2 graduate level courses on Bayesian modeling. I have now finished reading 6 chapters of the book as part of preparing for my candidacy exam and I am glad I took another shot at reading this book. I now understand fairly well issues with MCMC, what the limitations of Gibbs sampling are, when Metropolis-Hastings is typically used and why convergence diagnostics theory is hard. All of these are important practical issues for anybody who wishes to use these methods in their work.As other reviewers state, the writing can get dense sometimes. I skipped over some stuff that I felt was too much detail for my purposes and yet came away with a certain level of confidence about the knowledge I gained from the book. If you are ready to work out some of the derivations yourself and if you are willing to attempt to solve some of the exercises (which I did in my grad level class), you will get the most out of this book.So in conclusion, this is a must read for anybody serious about understanding Bayesian methods but this shouldn't be the first book you read on the topic. I ordered this book today because my library copy is overdue and I would like to have this book in my personal library.

"Throughout this book, you will find us fearlessly editorializing, telling you what you should and shouldn't do. We do not claim that our advice is infallible! Rather, we are reacting against a tendency, in the textbook literature of computation, to discuss every possible method that has ever been invented, without ever offering a practical judgement on relative merit" W.H.Press et al, Numerical Recipes 3rd Edition.Das Buch entstand aus Lectures Notes die der Autor in Portugisisch auf einer Tagung der Brasilianischen Statistischen Gesellschaft gehalten hat. Das Buch ist eine erweiterte Übersetzung dieser Notes. Es ist eine relativ umfassende Einführung in die Bayes'sche Methode und MCMC. Entsprechend der Entstehungsgeschichte setzt der Autor aber sehr gute allgemeine Kenntnisse in Wahrscheinlichkeitstheorie voraus. Der Sinn der Lecture Notes war offensichlich gestandenen Statistikern die moderne Bayes-Denkweise näher zu bringen. Ohne solide Vorkenntnisse ist das Buch keine Einführung sondern schlicht und einfach unverständlich.Das Buch leidet an Mutlosigkeit. Es gibt einen breiten Überblick über die Literatur, der Autor sagt aber so gut wie nie: Leitln vergessts des alles, machts es so. Wie bei vielen Vorlesungen weiss man nachher alles mögliche, aber man kann nichts wirklich. Das Buch ist sowohl inhaltlich als auch sprachlich wenig anregend. Auch wenn es von einem Brasilianer geschrieben wurde, würde ich es unter die Rubrik "Typisches Deutsches Professorenbuch" einreihen. Solide, aber so öd wie die Spielweise der Deutschen Fussball-Nationalmannschaft.Eine wesentliche bessere, viel kürzere und praktische Einführung - inklusive hochperformaten C++ Code - gibt es in Press et. al. Nachdem ich Press schon auf dem Bücherregal hatte, hätte ich mir dieses Buch sparen können. Press ist aber auch nicht perfekt. Unter Further Reading wird auch dieses Buch empfohlen.

Good book for sutudents learning stochastics.