Identification and estimation of the conditional average treatment effect with nonignorable missing covariates, treatment, and outcome

2026-02-22

AI Summary

This study addresses the challenge of unbiasedly estimating the conditional average treatment effect (CATE) in observational settings where covariates, treatment, and outcome variables may all be missing not at random (MNAR). The authors establish, for the first time, nonparametric identifiability of CATE under a multivariate MNAR mechanism and develop corresponding nonparametric and parametric estimation procedures. Furthermore, they introduce a sensitivity analysis framework to quantify the dependence of causal estimates on assumptions about the missingness mechanism. By overcoming the theoretical bottleneck in causal effect identification under complex missing data scenarios, this work provides empirical researchers with robust and verifiable inference tools that explicitly account for nonignorable missingness across multiple variables.

Technology Category

• Machine Learning: Causal Learning
• Reasoning under Uncertainty: Causality
• Cognitive Modeling & Cognitive Systems: Conceptual Inference and Reasoning

Application Category

• Graph Algorithms and Modeling for the Web: Algorithms and analysis for incomplete, noisy, or partially observed Web-related graphs
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Security and Privacy: Large-scale security measurements

Abstract

Treatment effect heterogeneity is central to policy evaluation, social science, and precision medicine, where interventions can affect individuals differently. In observational studies, covariates, treatment, and outcomes are often only partially observed. When missingness depends on unobserved values (missing not at random; MNAR), standard methods can yield biased estimates of the conditional average treatment effect (CATE). This paper establishes nonparametric identification of the CATE under multivariate MNAR mechanisms that allow covariates, treatment, and outcomes to be MNAR. It also develops nonparametric and parametric estimators and proposes a sensitivity analysis framework for assessing robustness to violations of the missingness assumptions.

Problem

• conditional average treatment effect
• nonignorable missingness
• missing not at random
• treatment effect heterogeneity
• observational studies

Innovation

• conditional average treatment effect
• nonignorable missingness
• missing not at random
• nonparametric identification
• sensitivity analysis