Different results with Stan 2.21.0

[wds15]

I think this went into Stan, but did not make it into the release notes (where it should have been).

Here is the respective git commit (which had a small bug which got fixed later)

Add additional no-u-turn checks, clearer naming conventions

here is the PR:

Feature/issue 2799 robust no u turn

and here is the long thread leading to this change

NUTS misses U-turns, runs in circles until max_treedepth

This is most likely related to the performance regression discussed by @bbbales2 and @betanalpha in the thread on stepsize adaptation tweak but not directly relevant for that thread’s topic. NUTS is supposed to automatically select the optimal integration time for HMC and because that should only depend on the geometry of the model I would expect that the number of leapfrog steps NUTS uses for a given model is inversely proportional to the stepsize. That is an overly simplistic argument but eve…

our update to rstan caused our production models to start spitting out different answers,

whow! Like the models give quite different answers? More details would be interesting… and you should ping people from the discourse thread I linked.