CausalEGM

Estimating Causal Effect by Deep Encoding Generative Modeling. CausalEGM utilizes deep generative neural newtworks for estimating the causal effect by decoupling the high-dimensional confounder into a set of different latent variables with specific dependency on treatment or potential outcome.

Requirements

• TensorFlow>=2.4.1
• Python>=3.6.1

Install

CausalEGM can be installed by

pip install causalEGM

Software has been tested on a Linux (Centos 7) with Python3.9. A GPU card is recommended for accelerating the training process.

Reproduction

This section provides instructions on how to reproduce results in the our paper.

Simulation data

We tested CausalEGM with simulation datasets first.