I went over this example again and in the lm full function they divide age by 10.
Thank you so much for any explanations and hints and apologies! I’m sure I"m missing something obvious once more, but I don’t see it.
Edit: Also, I find this a bit sparse: Generalized Linear Models — PyMC3 3.11.5 documentation is there more documentation to the different attributes and modules, arguments etc. somewhere?
I’m not sure you are missing something obvious. There’s mention that it helped with model convergence, so I guess that’s a clue. If I had to guess, I might speculate that including raw age was ok, but also including age^2 might make it somewhat difficult to use similar priors over the different coefficients. But that’s just a guess.
The GLM submodule is going away in v4 and users should migrate to bambi instead.
Thanks a lot!
The link isn’t working, can you check? Sorry
Edit: BAyesian Model-Building Interface (Bambi) in Python — Bambi 0.7.1 documentation I think this one
Yup, should be fixed now.
Thanks a lot bambi is awesome, so much easier!