Bob Carpenter

[Edit: this is wrong; see next message] To code this in Stan, you need to follow the approach taken for lognormal or inverse gamma. We don't directly wrap Boost implementations...

Thanks for reporting. > This is mathematically imprecise, as CDFs and PMFs should be defined for all real numbers (e.g., pbinom and dbinom in R accept non-integer values). Stan's definition...

Thanks for the elaboration. I'm still trying to imagine what the Stan code would look like. The obstacle to encoding binary, count, and real-valued entries is that Stan doesn't have...

The [Wikipedia entry for probability mass functions](https://en.wikipedia.org/wiki/Probability_mass_function) has two definitions in the first two paragraphs. Their simple definition matches @ethan-alt's (although restricts to real numbers, not any old `x`) and...

> (from @ethan-alt) Basically, the user would input, say, a N x 3 matrix of normal, Bernoulli, and Poisson dependent variables (along with covariates, if applicable) and have an indicator...

> (@spinkey): It's not "wrong" it's the "continuous" or "interpolated" version that doesn't use floor... Is it intended as a pdf rather than pmf, and if so, is it normalized?...

@bgoodri: Could you take a look at this function? Is it what you'd want for this problem. @spinkney: We can now declare functions where there is a data declaration so...

> figured it out Now I'm curious. I'm guessing the test uses more precision than the print. If that's the case, maybe we could change the way it prints by...

Approximating the true derivative will give you better precision w.r.t. the true value than differentiating the approximation of the value. But then the question is given that the approximation of...

OK. I invited you as a collaborator for this repo. You can always fork. This all looks great. My only concern is that we keep the math to a level...