The python package AutoDoubleML provides wrappers for double / debiased machine learning with automated nuisance function hyperparameter optimization. The package is built on top of DoubleML.
See a demo example at quickstart.ipynb.
Currently, only installation from GitHub is supported.
git clone https://github.com/OliverSchacht/AutoDoubleML
pip install requirements.txt
pip install .
For documentation of the double machine learning estimators please refer to the DoubleML documentation .
For information about hyperparameter tuning within double machine learning, see our 2024 CLeaR Paper.
AutoDoubleML is currently maintained by @OliverSchacht and @PhilippBach.
from autodml.AutoDoubleMLPLR import AutoDoubleMLPLR
from doubleml.datasets import make_plr_CCDDHNR2018
obj_dml_data = make_plr_CCDDHNR2018(alpha=0.5, n_obs=500, dim_x=20)
adml_plr = adml.AutoDoubleMLPLR(obj_dml_data, time=20)
adml_plr.fit().summary
coef std err t P>|t| 2.5 % 97.5 %
d 0.485355 0.041147 11.795644 4.110444e-32 0.404708 0.566001
Bugs can be reported to the issue tracker at https://github.com/OliverSchacht/AutoDoubleML/issues.
Contributions from the community are appreciated.