search

ZELLMECHANIK-DRESDEN / ShapeOut

Shape-Out has been superseded by Shape-Out 2.
https://shapeout.readthedocs.io
GNU General Public License v2.0
5 stars 5 forks source link

Linear Mixed Effects Models in Python instead of R? #144

Closed paulmueller closed 4 years ago

paulmueller commented 7 years ago

I think it would make sense to migrate the computation of linear mixed effects models from R to Python. The R installation seems to continuously make trouble, blows up the size of the Windows installer and makes it difficult to understand, test, and use the code. statsmodels seems to have this capability:

http://statsmodels.sourceforge.net/devel/mixed_linear.html

@MaikH88 Could statsmodels replace the current R workflow?

maikherbig commented 7 years ago

Finally linear mixed models are also available in statsmodels! I was able to reproduce the exact same results in lme4 and statsmodels for an example dataset using linear mixed models. Unfortunately generalized linear mixed models are not yet available. Therefore I would not yet switch. Still we could discuss dropping support for glmm: On the one hand it is nice that users can currently check their ignificance values at least using two models. On the other hand lmm and glmm give very similar results for us becasue lmm is already quite robust for non-normal data.

paulmueller commented 7 years ago

Thanks for the comment - good to know. I suggest we wait until glmm is in statsmodels. Are there any other opinions?

paulmueller commented 4 years ago

Unlikely that anything is going to change here, because the focus us on Shape-Out 2.