Markov chain Monte Carlo : stochastic simulation for Bayesian inference
Dani GamermanPreface
Almost a decade has elapsed since the release of the first edition. A large
amount o f recent work was produced on the MCMC subject but made no
substantial theoretical contribution. As anticipated in the first edition, most
of the ground work for the theory had been established by then. Subsequent
literature has basically enabled further understanding and extensions of the
previous work. In any case, the book has been updated to include the recent
literature and as a result the number of references has almost doubled. We
believe to have included at least a reference to most new developments in
MCMC.
What has really changed in this decade is the depth of understanding
and amount of applications of MCMC to the solution of inference problems.
The revision we performed concentrated on this point. The reader will
hopefully face a much more readable book in terms o f practical aspects. The
numbers of exercises, examples, numerical tables and figures have also been
considerably increased. We tried to exemplify and illustrate archetypical
situations to many applied areas to enable a better apprehension of the
pros and cons of the variety o f algorithms available in the MCMC arena.
In line with the modern resources available nowadays, the URL site
www.ufrj.br/MCMC has been created. It contains the codes (all written
in R language) used in many of the previously existing and new examples
and exercises of the book. Readers will have free access to them and will
be able to reproduce the tables and figures of the book. More importantly,
the mildly self-explanatory nature of the codes will enable modification of
the inputs to the codes and variation in many directions will be available
for further exploration. This internet tool is planned to be constantly being
updated and can also be used to compensate for any new development not
included in this edition of the book.
The major changes from the previous edition are as follows. New sections
on spatial models and model adequacy have been introduced in Chapter
2. Spatial models is an area that has experienced a huge development in
statistics during the last decade and the writers of the book have made
a few contributions there as well. A section on model adequacy should
have always been there. All that was done was to minimally remedy this
flaw of the first edition. Chapter 7 is the chapter that has undergone the
largest change. It has moved away from its speculative flavor to a much
more detailed description of a number o f techniques that are routinely
used nowadays. Chapters 5 and 6 have also been considerably increased
by inclusion of more illustrative material. This was done with the sole
aim o f providing better understanding of the MCMC machinery. All other
chapters have been subjected to additions but to a smaller amount.
In summary, the book has been substantially reinforced as a first reading
material on MCMC and, consequently, as a textbook on modern Bayesian
computation and Bayesian inference courses. More advanced derivations
were not present before and are still not present in this edition.