This is the codebase for the paper "Conformal Meta-learners for Predictive Inference of Individual Treatment Effects". It includes the implementation of a general framework for issuing predictive intervals for individual treatment effects (ITEs) by applying the standard conformal prediction (CP) procedure on top of pseudo-outcome regression models of the conditional average treatment effects (CATEs). You can use this codebase to apply conformal meta-learners to new datasets or reproduce the experiments in our paper.
Download the codebase from source and install all dependencies in requirements.txt.
To execute all experiments in the paper, you can run the following shell script
bash experiments.sh
To change the parameters of an individual experiment, you can use the following command
python run_conformal_metalearners.py -t "test fracion" -b "List of baselines" \
-s "Synthetic data setup" -e "Dataset type" \
-n "Number of synthetic data points" \
-d "Number of feature dimensions" -q "Use of quantile regression"\
-v "Saving figures" -x "Number of experiments" \
-c "Target coverage" -w "Sweep all coverage probabilities"
The list of meta-learners baselines is ["X", "DR", "IPW"]. The parameter -s can be set to "A" or "B" to select the synthetic data setup in Section 5.1. The parameter -e designates the dataset type, with three options: "Synthetic", "IHDP" and "NLSM". The number of data points and feature dimensions are ignored in real data experiments. The parameter -q determines if the meta-learners use quantile regression in the base learners (mean regression is used if -q is set to False). Results of experiments are saved in the figures and log folders.
If you use our code in your research, please cite:
@article{alaa2023conformal,
title={Conformal Meta-learners for Predictive Inference of Individual Treatment Effects},
author={Alaa, Ahmed and Ahmad, Zaid and van der Laan, Mark},
journal={Advances in Neural Information Processing Systems},
year={2023}
}