search

fabsig / GPBoost

Combining tree-boosting with Gaussian process and mixed effects models
Other
575 stars 46 forks source link

tests / p-values for groups of multiple variables in GLMMs #81

Open qiuyugong-aifi opened 1 year ago

qiuyugong-aifi commented 1 year ago

Hi,

Thank you for this great package.

I am just wondering is there a way to get p value of the model like following? Screen Shot 2022-12-12 at 12 15 15 PM Screen Shot 2022-12-12 at 12 15 36 PM

Thank you!

Q

fabsig commented 1 year ago

By running the code as you do it you obtain p-values for individual variables. Anova-style p-values for multiple variables (e.g., all fixed effects dummies of a factorial variable) is currently not possible.

While being simple from a statistical point of view, the question is how to implement this consistently in both the R and the Python package. What is currently not yet entirely clear to me is how do we let the software (automatically) know which variables belong together (given that GPBoost does not use the R-style formula notation). I will think about it and let you know if there is an update.