Christian Lorentzen

@TimotheeMathieu Thanks for the explanation. Now I get it. Something that could be mentioned in the example is the trimmed mean as a simpler entry point to robust estimation.

As info, patsy has issues with pickle, see https://github.com/pydata/patsy/issues/26.

Why not ```python y, X, model_spec = model_matrix("y ~ a + b + a:b", df) ``` This is also clean and avoids patching data containers. If one attaches *metadata* like...

Is the above list up to date? I'm curious if I could be of any help.

Hope you and your family are fine. In my opinion, `C` is the most important of all. I'll have a look at the spline basis, let's see.

I'd be interested in your opinions about `C`.

IIUC, this would mean that we define a *plugin API*. Is this correct? What is the advantage of such a plugin mechanism vs libraries implementing their own estimators?

If I may, I'd like to raise some stupid questions: 1. Do we need to change our statement on GPU in the [FAQ](https://scikit-learn.org/stable/faq.html?highlight=gpu#will-you-add-gpu-support)? GPU's are one of the main motivations...

> >What happens with 3rd party transformers in a pipeline if global set_config(transform_output="pandas") has been set? >If they inherit from TransformerMixin, then they will auto opt-into set_output and transform will...

@thomasjpfan The only open point for me is the name of the default option of `transform_output`, see https://github.com/scikit-learn/scikit-learn/pull/23734#discussion_r985071192. Your suggestion to use `None` instead of `"default"` is good for me....