Carlos Parada

> (FWIW, in my field the BIC is also often abused to claim that a variable has no effect, yet I wouldn't suggest dropping it as it's a classic indicator.)...

> > A higher R^2 indicates a better model > > No, even assuming that there is a clear definition of R^2 for a given model (which there isn't for...

> Again, no! For one thing, R2 captures a particular measure of variance (relative to a saturated model) and not the variance of the predictions. The Stein paradox and the...

> And I don't think we want to in any specialize on Tables.jl in Turing.jl (this is just my personal opinion though!). This too should just be wrapped in something...

> @ParadaCarleton Something like this: [TuringLang/AbstractPPL.jl#26](https://github.com/TuringLang/AbstractPPL.jl/pull/26)? I have no idea what that's doing but if you claim it helps I'll believe you. :sweat_smile:

> Sounds like a good idea. PR welcome 😉 Working on this ATM, which optimization package do you recommend for optimizing a continuous parameter? I know there's a dozen implementations...

> I see in other threads you have been considering using MLJTuning.jl for optimization. I think this is probably overkill, and will require you to formulate the problem as a...

I think what you're suggesting works, but for now I've implemented a hard-coded optimizer using Newton's method :sweat_smile: Mostly just to separate this into 2 PRs, since overhauling the whole...

This would be amazing to have--any more progress?

Side note for the future--besides documenting this, there's better rules than the Sturges rule nowadays, which tends to suggest far too few bins. More modern rules suggest `O(n^(1/3))` bins. I...