Bob Carpenter

@JackCaster: Thanks. If @avehtari's definitions are for the parameterization you want, could you please indicate that in the issue itself? Then the question is what to call the distribution. I...

Thanks, @avehtari. For naming, it's more for consistency---our other distribution names are `inv_gamma`, `neg_binomial`, etc. But I also think it's clearer if someone doesn't know the gpd TLA.

Sorry, @JackCaster, but we gave up on writing roadmaps or milestones because they were too contentious among the developers. So now we just try to triage features into ones that...

Is the only use case you anticipate through the Stan language? Is the idea to have approximate distributions, like poisson_approx or something? This issue eventually needs to specify what the...

> Ripping code apart into testable things is not at all not smart from what I learned. There are at least three concerns that need to be balanced, 1. efficiency...

> It turns out that we need more precision when doing the forward pass during reverse mode which I found kind of weird and interesting The forward-mode values are often...

Thanks for the correction, @nhuure---you're absolutely right. It's too bad we have to waste a parameter on it, but I suppose we have to for uniformity. It'll just be a...

Is this the intended definition, or if not, could you clarify. ``` lognormal_glm_lpdf(y, x, alpha, beta, sigma) = normal_glm_lpdf(log(y), x, alpha, beta, sigma) = lognormal_lpdf(y | exp(alpha + x *...

Anything that requires a Jacobian is going to be tricky for a lot of our users. Also, I strongly prefer not to have transforms everywhere in code. So I think...

@spinkney: Could you provide an example and ideally the signatures of operators or functions you'd like to add? Given that we do allow increments and decrements of array elements, I'm...