Disentangled Representation Learning for Causal Inference with Instruments

TL;DR

This paper introduces a VAE-based method to learn instrumental variable representations from observational data with latent confounders, enabling unbiased causal effect estimation without requiring known IVs.

Contribution

It proposes a novel disentangled representation learning approach that relaxes the need for known IVs, broadening the applicability of IV-based causal inference methods.

Findings

Outperforms existing IV estimators on synthetic data

Achieves more accurate causal estimates on real-world data

Effectively learns IV representations without prior knowledge of proxies

Abstract

Latent confounders are a fundamental challenge for inferring causal effects from observational data. The instrumental variable (IV) approach is a practical way to address this challenge. Existing IV based estimators need a known IV or other strong assumptions, such as the existence of two or more IVs in the system, which limits the application of the IV approach. In this paper, we consider a relaxed requirement, which assumes there is an IV proxy in the system without knowing which variable is the proxy. We propose a Variational AutoEncoder (VAE) based disentangled representation learning method to learn an IV representation from a dataset with latent confounders and then utilise the IV representation to obtain an unbiased estimation of the causal effect from the data. Extensive experiments on synthetic and real-world data have demonstrated that the proposed algorithm outperforms the existing IV based estimators and VAE-based estimators.