Proximal Causal Learning under Unmeasured Confounding

2026-10-03

AI Summary

This study addresses the bias in treatment effect estimation caused by unmeasured confounding in observational data, as well as the limitation of existing methods that rely on predefined proxy variables. To overcome these challenges, this work proposes PCL-U, an end-to-end proximal causal learning framework. The method employs a neural encoder to automatically decompose covariates into treatment-inducing, outcome-inducing, and shared proxy variables, and integrates a minimax mutual information optimization objective with a moment risk function for causal inference. Experimental results demonstrate that PCL-U matches or surpasses existing baselines on benchmark datasets while maintaining stable accuracy across various synthetic data settings. Ultimately, the proposed framework effectively mitigates estimation bias arising from unmeasured confounding.

Abstract

Estimating treatment effects from observational data typically relies on the No Unmeasured Confounding Assumption (NUCA), which rarely holds in practice. Proximal causal learning (PCL) addresses unmeasured confounding via proxy variables, yet existing methods require the proxy variables to be pre-specified. Thus, we propose PCL-U, a framework that learns proxy variables directly from observed covariates. PCL-U uses neural encoders to decompose covariates into treatment-inducing, outcome-inducing, and shared proxies, guided by minimax mutual information objectives, and obtains causal estimates through a practical moment-based risk function. Experiments on benchmarks show that PCL-U matches or outperforms existing baselines. Besides, there are two types of synthetic datasets with varying dimensions and confounding strengths that illustrate that our method maintains stable estimation accuracy.

Problem

• Proximal Causal Learning
• Unmeasured Confounding
• Treatment Effect Estimation
• Proxy Variables
• Observational Data

Innovation

• Proximal Causal Learning
• Unmeasured Confounding
• Proxy Variables
• Neural Encoders
• Minimax Mutual Information