Radu Craiu, Dootika Vats, Galin Jones, Steve Brooks, Xiao-Li Meng, and I edited the second edition of the Handbook of Markov chain Monte Carlo. Dootika set up a github page for the book, listing all the chapters and includes links to most of them in Arxiv form. (Chapter 4, “For how many iterations should we run Markov chain Monte Carlo?”, is by Charles Margossian and me.) For some reason, a few of the chapters are not yet on Arxiv but I guess they’ll get there soon.
I recommend the whole book, but especially chapter 24, “Running Markov chain Monte Carlo on modern hardware and software,” by Pavel Sountsov, Colin Carroll, and Matthew Hoffman. But really, just dive in and read whatever chapters interest you.
My only regrets are that we didn’t include chapters on the following topics:
– Probabilistic programming (Stan, etc.)
– Sequential Monte Carlo (particle filtering)
– Divide-and-conquer algorithms (expectation propagation, etc.)
But, hey, no project is ever done.
I’m glad to have been part of this, and special thanks to Radu and Dootika, who joined the project for the second edition and added a lot.
Thanks for sharing this resource! Between this and the updated User’s Guide, you’re providing a lot of helpful resources for those learning to do Bayesian data analysis. Most of the students I work with can handle the theory but stumble over the technical details of model implementation and computation.
To that end, I want to thank you folks also for the Stan Playground, another extremely helpful resource for newcomers. I just used it to prepare an example illustrating a complete analysis pipeline for a project a student is working on. It only took about an hour for me to prepare the complete worked example and now all I have to do is share a ZIP file and the student can start tinkering with it themselves.
Oh, yeah, I loooove Stan Playground.
GitHub pages probably would have been the more appropriate tool here. That just renders as a web page.
The original web page for the book is now blank if anyone’s monitoring it—it’s still one of the first hits on the web when you search:
https://www.mcmchandbook.net/
Radu’s surname (“Craiu”) was misspelled. I fixed it in the post and submitted an issue for the GitHub that included a link to Neal’s, Geyer’s, and Fearnhead/Sherlock’s papers. In case you want the links now, here’s the issue:
https://github.com/dvats/HandbookOfMCMC/issues
Thanks!
Thank you, Bob!
Thank you. This is really useful. I kindly suggest a bibtex file with the correct way to cite these chapters.
I’ve found Geyer’s first chapter particularly useful. It isn’t on the arXiv, though it’s out there online as Bob’s noticed. If copyright permitted, I kindly suggest you host a version of it in the repo, as the online links may disappear.
According to Charles Margossian, Geyer revised the chapter to include a criticism of the many short chains regime (and probably other changes, but that’s all that Charles mentioned). So the editors probably wouldn’t want to link the old one for the new edition. I’ve never seen a second edition of a collection like this before that revised the papers and even changed the author lists.
Aside from the polemics, I found Geyer’s article useful. The intro to MCMC that I find most useful and recommend the most is the one by Roberts and Rosenthal (2004): General state space Markov chains and MCMC algorithms.
Bob:
This second edition is an entirely new book. If there are any chapters unchanged from the old edition, there are only one or two. The field has just changed so much.
There is a chapter on MCMC for State Space Models by Fearnhead and Sherlock that includes a discussion of Particle filtering.
Would you have a reference for the “Divide-and-conquer algorithms” mentioned at the end of the post?
There’s a whole literature on this. Here are a few papers I’ve been involved in:
– [2017] Consensus Monte Carlo using expectation propagation
– [2020] Expectation propagation as a way of life: A framework for Bayesian inference on partitioned data
– [2023] Federated learning as variational inference: A scalable expectation propagation approach