Looking at the documentation for pymc3.distributions.bound.Bound, I see this sentence:
The bounds are inclusive for discrete distributions.
I assume, because this is mentioned as a contrast, that the bounds are not inclusive (they are strict) for continuous distributions, but I couldn’t find any clear statement to this effect. Is this true?
I suppose mathematically, the probability of hitting exactly the bound is zero, but since we are doing computer approximations, I wanted to make sure that the sampler could never hit the bound.
I’ll try to make a PR for a documentation fix when I get an answer.
Thanks!