Which models to include?

[TimTaylor]

Currently lm, glm, MASS::glm.nb, brms::brm

[dirkschumacher]

It would be interesting to use tree based models. Ideally something like xgboost. We need to find out if there exist proper prediction intervals.

[dirkschumacher]

In addition I would suggest glmnet. Anything that can do regularized regression

[TimTaylor]

Sounds good. Not used glmnet but it looks like we will probably need glmnetutils for a the formula interface.

[dirkschumacher]

Oh yeah, that is probably necessary, as they have a pure matrix based interface

[TimTaylor]

Also, not that it's a problem, I don't think we could add confidence/prediction intervals for glmnet.

[dirkschumacher]

Yeah, of course we need to research proper prediction intervals for regularized regression.