causal_multitask_gaussian_processes_ite

Bayesian Inference of Individualized Treatment Effects using Multi-task Gaussian Processes

Dependencies

This implementation requires the IHDP dataset introduced in [3].

Usage

python test_models.py -n "number of experiments" -t "test data ratio" -m "mode" [ -o <result.json> ]

The argument "mode" can be set to "NSGP" or "CMGP".

References

1. Bayesian Inference of Individualized Treatment Effects using Multi-task Gaussian Processes
2. Limits of Estimating Heterogeneous Treatment Effects:Guidelines for Practical Algorithm Design
3. J. L. Hill. Bayesian Nonparametric Modeling for Causal Inference. Journal of Computational and Graphical Statistics, 2012.