Luca Bittarello

> Another question is if one should ever rely on the order of categories in Pandas categorical types... Perhaps be explicit about it and cast to ordered categorical in `transform`?

I don't know any Cython. :T @xhochy?

> Should we define the more basic `__mul__` and `__rmul__`? Yes, please! With a consistent API across subtypes. At the moment, `*` performs element-wise multiplication for dense matrices (NumPy style),...

To my mind, it's okay if one subtype has a fuller API than the others. I only mentioned `__mul__` (and indexing elsewhere) because it seems pretty elementary to me. It...

What should the interface look like? `fit` is pretty straightforward, but I'm not sure about `predict`.

> A final question is [...] whether some of the work we did here could be extended provide a basis for a GAM library There are different ways to fit...

Here's the rationale for `score` and `jacobian`: The asymptotic variance of GMM estimators takes the form E (ψ Ω ψ'), where ψ is the influence function and Ω is the...

Very late to the game, but: Why do we use a pseudo-R² as the score at all? Couldn't we just report the negative or inverse deviance? For tuning purposes, it...

Closed as unresponsive. Feel free to reopen!

Glum now offers a `drop_first` option (#571).