Orthogonal Statistical Learning

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/FLAML
• econml: https://github.com/microsoft/EconML
• scikit-learn: https://github.com/scikit-learn/scikit-learn
• numpy: 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 models
• cate.py: contains implementations of all the CATE estimation methods
• policy.py: contains implmentations of all the Policy learning methods
• slearner.py: contains an SLearner cate estimator class that is used in cate.py.