ParallelGSReg / GlobalSearchRegression.jl

Julia's HPC command for automatic feature/model selection using all-subset-regression approaches
Other
18 stars 3 forks source link

No intercept only model #7

Open EvoArt opened 4 years ago

EvoArt commented 4 years ago

First, thanks for this brilliant function!

Perhaps my issue is due to my own lack of expertise. However, the standard practise for model averaging (looking at the literature and "dredge" in R/MuMln) seems to be to include a null/intercept only model in the model comparisons and potentially the averaged model. In order to achieve this with GSReg, I have tried using intercept = false and adding an explanatory variable that is just a column of ones. Of course this results in a models without intercepts being considered.

Is there any way around this?