Daniel Lee

@bgoodri, do you envision this as a method on the Stan program class? or something that's reported by the sampler? Now that I think about it, it seems like it...

We could use inversion of control to have the interfaces, at runtime, choose what they want back. It doesn't have to affect the implementation of the algorithms at all. >...

(This doesn't solve rhat due to fit, but no reason we can't do it for means and variances.) > On Jul 25, 2017, at 10:17 AM, Daniel Lee wrote: >...

Exactly. But now I realize we don't even need to expose additional callbacks. We could use a visitor or chain of responsibilities pattern. As I'm working on the better c++...

Add one more vote for no iteration number within the language. If you're really inclined, there are ways to use that information from an algorithm written in C++. You might...

Good points. Especially about the prototype and it's a developer-level problem. If it's just about printing, it seems like a lot of effort. Probably worth it in the long run...

@steven-murray, can you provide a more concrete description of the incomplete gamma function you're looking for? And, is this for use in the Stan language? Or just in C++.

@ariddell, I think this is a language issue. @bgoodri, are you envisioning something like this only happening in generated quantities?

Hmm... that makes it much more of a pain, but I don't think it's impossible. We'd need to add the `score()` function to the math library that does the nested...

Have any suggestions on how to update golden records? I'd rather have difficult tests over breaking things inadvertently. This comes from having things being broken multiple times through the history...