Vectorise `formula` in model functions
- 1 formula repeated for each key
-
ndifferent formulas (lists of formulas under the hood, we could discuss a more user-friendly constructor later), wheren = # of key.
Recycling rules can be implemented in fablelite.
The modelr::formulas() API looks very useful. https://modelr.tidyverse.org/reference/formulas.html
Yep - looks like that's suitable for our usage. I'd probably prefer the RHS to be bare, as this is more consistent with our support for bare formulae being the response.
I've thought about this some more, and I prefer introducing a model_when() function, rather than vectorising the formula. This will allow you to apply different model specifications (and accordingly, different model formulae) to different sections of the data.
It may also be possible to use case_when() rather than introducing a new function.
Is there still no way to specify different model formulas for different time series inside model()?