Stan Meeting Jan 24, 2019


#1

Andrew: Lots of ideas the past week. Nothing concrete.

Lauren: Working with Andrew, talking to folks in Australia. How does anxiety and exercise relate? Looking into evaluation scales for common psy metrics.

Matthijs: Gave a talk, Marie gave a talk. 400 folks or so. Met with Flatiron institute.

Breck: StanCon is up, got a funding proposal out.

Sean: New Stan compiler, met with IBM research folks that do YAPS. Working with Sebastian on threading.

Bob: Wrote first OCaml program. Google grant out, Northern Az grant out.

Jonah: Mitzi, Advait conversation about Stan book. Fixing RStanArm stuff.

Paul: BRMS bug fixes. Working on paper with Aki on cross validation for time-series models, revisions on other papers.

Sebastian: Map_rect was not working on Windows, now it is. PR being reviewed. Can we improve sharding or blocking?

Aki: Papers on their way out.

Mitzi: Met with IBM folks about PyCmdStan. Worked on standalone generated quantities.


General discussion:

Do we want to have backwards compatible math API? We generally don’t do this, need to update policy.

What to do about new thread initialization requirement? This will help Stan scale when there are a lot of threads to work with. Should work even for single cpu if there is a lot of data.

Discussion for PK/PD models on individual level. Is this currently done, is Bayesian approach correct?