Experiments for the paper: Orthogonal Statistical Learning
To generate all paper tables run the jupyter notebook RunAllExperiments.ipynb
The code requires the following python packages:
flaml
: https://github.com/microsoft/FLAMLeconml
: https://github.com/microsoft/EconMLscikit-learn
: https://github.com/scikit-learn/scikit-learnnumpy
: https://numpy.org/pandas
: https://github.com/pandas-dev/pandas/
The code is organized as follows:
experiments.py
: Contains all the logic for generating data for different setups, running an experiment and storing the result.automl.py
: contains wrappers for nuisance and target automl FLAML modelscate.py
: contains implementations of all the CATE estimation methodspolicy.py
: contains implmentations of all the Policy learning methodsslearner.py
: contains an SLearner cate estimator class that is used incate.py
.