Conditional Distributional Treatment Effects: Doubly Robust Estimation and Testing

2026-03-17

AI Summary

Existing methods struggle to characterize the heterogeneous effects of treatment on the full outcome distribution—such as variance and tail behavior—under covariate dependence, and lack a theoretically grounded global test for distributional homogeneity. This work proposes a novel estimator for conditional distributional treatment effects and develops a locally asymptotically minimax optimal doubly robust estimator. Furthermore, it introduces a computationally efficient, permutation-free global test that, for the first time in this setting, guarantees strict Type I error control and consistency against fixed alternatives. The approach also yields closed-form solutions for distributional discrepancy measures such as Maximum Mean Discrepancy (MMD), achieving both statistical efficiency and computational tractability.

Technology Category

• Machine Learning: Calibration & Uncertainty Quantification
• Constraint Satisfaction and Optimization: Distributed CSP/Optimization
• Search and Optimization: Distributed Search

Application Category

• Web Mining and Content Analysis: Robustness and generalizability of Web mining methods
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Responsible Web: Human-perceived consequences of algorithmic deployment on the web

Abstract

Beyond conditional average treatment effects, treatments may impact the entire outcome distribution in covariate-dependent ways, for example, by altering the variance or tail risks for specific subpopulations. We propose a novel estimand to capture such conditional distributional treatment effects, and develop a doubly robust estimator that is minimax optimal in the local asymptotic sense. Using this, we develop a test for the global homogeneity of conditional potential outcome distributions that accommodates discrepancies beyond the maximum mean discrepancy (MMD), has provably valid type 1 error, and is consistent against fixed alternatives -- the first test, to our knowledge, with such guarantees in this setting. Furthermore, we derive exact closed-form expressions for two natural discrepancies (including the MMD), and provide a computationally efficient, permutation-free algorithm for our test.

Problem

• conditional distributional treatment effects
• potential outcome distributions
• heterogeneous treatment effects
• distributional discrepancies
• causal inference

Innovation

• Conditional Distributional Treatment Effects
• Doubly Robust Estimation
• Minimax Optimality
• Distributional Homogeneity Test
• Maximum Mean Discrepancy