Have any tests been done to see which of the many interfaces to Stan is fastest? How does this differ per data size and complexity of the model?
I would guess that the fastest way to run things would be directly in CmdStan.
Kind regards
Have any tests been done to see which of the many interfaces to Stan is fastest? How does this differ per data size and complexity of the model?
I would guess that the fastest way to run things would be directly in CmdStan.
Kind regards
It is absolutely negligible compared to the time needed to draw from the posterior distribution.