Stan this Month
Issue 2, February 2019
Editors: Breck Baldwin, Charles Margossian
Stan this month is a newsletter aimed at highlighting major developments, discussions, and ideas within the Stan community. It provides an overview and pointers to more detailed references. If you wish to subscribe to this newsletter click here.
This month’s issue covers:
- Resources for beginners learning Stan
- Announcing the new director of the Technical Working Group (TWG)
- Soliciting interested people for a user meeting in New York City to help with the Stan roadmap
Resources for people new to Stan
Stan can be quite challenging to learn. Depending on your background you may have to learn any of:
- Statistics and probability theory
- Programming in an interface language (such as R and Python) and in Stan
- Bayesian modeling
This month, we share useful links and information to help get you started. A great place to ask questions, whether you are fitting your first model, proposing new features for the language, or anything in between, is the discussion forum on discourse: https://discourse.mc-stan.org/.
For a first glance, there is a web accessible version of the R interface to Stan at RStudio Cloud. Look at simple_regression.html in the files tab for some instructions–thanks to RStudio for creating this resource.
The top level page for users is at: https://mc-stan.org/users/ which will likely remain stable for the foreseeable future. That page has links to a page for installing interface languages to Stan and Stan itself. In addition the documentation page contains the majority of the links you will want to explore and we will not directly link any further since it is worth while exploring all that is on the page. The Stan user’s guide is an excellent way to get started if you know statistics. All paths to Stan mastery go through the user’s guide. The tutorial page can be a bit overwhelming with options but it covers books, online texts, slides and videos of varying lengths and topics. Case studies are also quite varied and helpful if your interest aligns with one of them.
For those who are not formally trained in statistics, two books are quite helpful:
- John Kruschke (2015) Doing Bayesian Data Analysis, Second Edition: A Tutorial with R, JAGS, and Stan
- Richard McElreath (2016) Statistical Rethinking: A Bayesian Course with Examples in R and Stan.
Kruschke presupposes very little math beyond algebra and spends time explaining how sampling works, McElreath assumes a bit more math, has less about the underlying mechanics but really gets the point across about what Bayesian analysis is.
Additionally, Bob Carpenter’s “Pooling with Hierarchical Models for Repeated Binary Trials” case study is a very clear example that works from a simple single parameter model to a hierarchical model with working code. Michael Betancourt has written several case studies, available on his website. These are quite demanding, but build the theory from the ground up, starting with probability theory, discussing algorithmic and computational considerations, and then diving into scientific modeling. Bill Gillespie has published a series of introductory lectures on Stan, which serve as a primer to his more advanced course on Bayesian modeling in pharmacometrics (i.e. quantitative pharmacology).
Next, here are two papers I (Charles) like to share with students when I teach an introductory workshop. We don’t discuss them during the class, rather it’s something for them to read if they want to go a little further. Philosophy and the practice of Bayesian analysis (A. Gelman and C. Shalizi., 2013) is a conceptual article that discusses various practices of Bayesian statistics, and where these fall with respect to the scientific paradigm. The second paper is for those who want to crack open the black box and learn more about one of Stan’s foundational algorithm. A Conceptual introduction to Hamiltonian Monte Carlo(M. Betancourt, 2018) provides many theoretical insights, which ultimately inform applications of Stan.
Sean Talts is the new TWG director.
The Stan Governing Body (SGB) elected Sean Talts to lead the Technical Working Group. Please look at the announcement on Discourse for more details, what his role is and how all this happened.
Call for interest in a users meeting to help set the technical roadmap for Stan
Stan remains a young project with many directions it could go. We would like to have a face-to-face forum for discussing how we can best serve our users. The meeting will help prioritize development and allocate resources.
We have been thinking that a one to two day meeting in New York City would be best, hopefully in the next month or two. This is very preliminary and all we are asking for at this point is an expression of interest. If you are interested please sign up on the dedicated mailing list for the meeting. The mailing list will only be used for updates about this meeting.
If you have comments, wish to contribute or have other ideas send email to stan-this-month@mc-stan.org.
See you next month.
If you wish to subscribe to this newsletter click here.