Add Target Scaling

[glevv]

I think it will have huge impact on results and on model training. I think pymc-marketing does it too.

Right now you can do standardize(y) in formula, but it does not support inverse transformations, which is inconvenient for result analysis and plotting.

[drbenvincent]

Hi @glevv. Sorry for the delayed response. I think this is a good idea. I assume you're after some kind of kwarg which automatically does this behind the scenes and uses something like https://scikit-learn.org/stable/modules/generated/sklearn.preprocessing.StandardScaler.html#sklearn.preprocessing.StandardScaler ?

Is this something you'd like to take a stab at? Either way, marking this as a feature request

[glevv]

Yes, something like this. An argument that will use StandardScaler on target variable and inverse transformation on the prediction, so that PyMC models would fit properly.
I'm not sure how to add this functionality to the library.