Hey Stan Devs,
I’m curious if you can impart some high level wisdom about testing probabilistic software. I have two questions:
- How do you test code that samples random variables?
I pretty much have no clue where to start with this, and am not quite C++ fluent enough to learn by reading the Stan source (also I’m not sure what part of the source to look at).
- How do you test for performance regressions?
I’m re-writing some of the slow R portions of the samplers in C++, but want to be sure that this actually results in faster code.
Figured y’all would be good people to ask about this kind of thing.