Scikit-learn estimator wrappers for pymer4 wrapped LME4 mixed effects models
sklearn-lmer - Scikit-learn estimator wrappers for pymer4 wrapped LME4 mixed effects models
Refer to the documentation for examples and api.
Linear mixed effects regressions
Linear mixed effects regressions are great, but if you’re here, you probably already agree. You can find more infomration about them elsewhere, the links lme4 aren’t a bad place to start.
Mixing r and python used to be a bit more fraught, but rpy2 and conda seem to be working together better these days. To install first get a conda environment with the dependencies:
>>> conda create -n sklmer numpy scipy rpy2=2.9.4 r-lme4 r-lmertest r-lsmeans tzlocal
Then pip install sklearn-lmer:
>>> pip install sklearn-lmer
It can be imported as:
>>> from sklmer import LmerRegressor
Now the mostly part of that compatiblity is that init does have two required paramters: a formula and the names of the columns holding independent variables and grouping variables (I’ve called this parameter X_cols even though it is more than just X). When I use this I’ve got my data in a dataframe and just pass dataframe.columns with X_cols like so
>>> df = pd.read_csv(os.path.join(get_resource_path(),'sample_data.csv')) >>> lreg = LmerRegressor('DV ~ IV2 + (IV2|Group)', X_cols=df.columns)
If you want the best compatibility with sklearn it probably makes sense to split out the dataframe into X, y, and group variables, though since you’ve defined a formula it’s ok if the y and group columns are in X
>>> X = df.values >>> y = df.DV.values >>> groups = df.Group.values
Once you’ve done that, it seems to work fine with other sklearn tools, like cross_val_score
>>> logo = LeaveOneGroupOut() >>> cross_val_score(lreg, X=X, y=y, cv=logo.split(X, groups=groups), scoring='neg_mean_squared_error')
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
|Filename, size||File type||Python version||Upload date||Hashes|
|Filename, size sklearn_lmer-0.0.4-py3-none-any.whl (9.1 kB)||File type Wheel||Python version py3||Upload date||Hashes View hashes|
|Filename, size sklearn-lmer-0.0.4.tar.gz (21.4 kB)||File type Source||Python version None||Upload date||Hashes View hashes|
Hashes for sklearn_lmer-0.0.4-py3-none-any.whl