# Unveiling the Potential of Robustness in Selecting Conditional Average Treatment Effect Estimators

Source: [Unveiling the Potential of Robustness in Selecting Conditional Average Treatment Effect Estimators](<https://www.sota2.com/research/paper/68fa75751e209131cea680d3>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2026-01-02T07:46:41.844Z

Authors: Yiyan Huang, Cheuk Hang Leung, Siyi Wang, Yijun Li, Qi Wu

Published: 2024-02-28T15:12:24.000Z

## Abstract

The growing demand for personalized decision-making has led to a surge of interest in estimating the Conditional Average Treatment Effect (CATE). Various types of CATE estimators have been developed with advancements in machine learning and causal inference. However, selecting the desirable CATE estimator through a conventional model validation procedure remains impractical due to the absence of counterfactual outcomes in observational data. Existing approaches for CATE estimator selection, such as plug-in and pseudo-outcome metrics, face two challenges. First, they must determine the metric form and the underlying machine learning models for fitting nuisance parameters (e.g., outcome function, propensity function, and plug-in learner). Second, they lack a specific focus on selecting a robust CATE estimator. To address these challenges, this paper introduces a Distributionally Robust Metric (DRM) for CATE estimator selection. The proposed DRM is nuisance-free, eliminating the need to fit models for nuisance parameters, and it effectively prioritizes the selection of a distributionally robust CATE estimator. The experimental results validate the effectiveness of the DRM method in selecting CATE estimators that are robust to the distribution shift incurred by covariate shift and hidden confounders.

## Recorded benchmark results

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

- [CATE Estimator Selection on ACIC Setting A (p=0) 2016](<https://www.sota2.com/research/sota/cate-estimator-selection-on-acic-setting-a-p-0-2016.md>): CATE Estimator Selection; ACIC Setting A (p=0) 2016; PEHE: 2.68.

- [CATE Estimator Selection on ACIC Setting A (p=0.1) 2016](<https://www.sota2.com/research/sota/cate-estimator-selection-on-acic-setting-a-p-0-1-2016.md>): CATE Estimator Selection; ACIC Setting A (p=0.1) 2016; PEHE: 3.55.

- [CATE Estimator Selection on ACIC Setting A, p=0.3 2016](<https://www.sota2.com/research/sota/cate-estimator-selection-on-acic-setting-a-p-0-3-2016.md>): CATE Estimator Selection; ACIC Setting A (p=0.3) 2016; PEHE: 5.28.

- [CATE Estimator Selection on ACIC Setting C, m=0.1 2016](<https://www.sota2.com/research/sota/cate-estimator-selection-on-acic-setting-c-m-0-1-2016.md>): CATE Estimator Selection; ACIC Setting C (m=0.1) 2016; PEHE: 4.6.

- [CATE Estimator Selection on ACIC Setting C, m=0.5 2016](<https://www.sota2.com/research/sota/cate-estimator-selection-on-acic-setting-c-m-0-5-2016.md>): CATE Estimator Selection; ACIC Setting C (m=0.5) 2016; PEHE: 6.44.

- [CATE Estimator Selection on ACIC2016 Setting C m=0.9](<https://www.sota2.com/research/sota/cate-estimator-selection-on-acic2016-setting-c-m-0-9.md>): CATE Estimator Selection; ACIC2016 Setting C m=0.9; PEHE: 10.05.

- [CATE Estimator Selection on ACIC Setting B, xi=0 2016](<https://www.sota2.com/research/sota/cate-estimator-selection-on-acic-setting-b-xi-0-2016.md>): CATE Estimator Selection; ACIC Setting B (xi=0) 2016; PEHE: 2.14.

- [CATE Estimator Ranking on ACIC Setting A (p=0) 2016](<https://www.sota2.com/research/sota/cate-estimator-ranking-on-acic-setting-a-p-0-2016.md>): CATE Estimator Ranking; ACIC Setting A (p=0) 2016; Rank Correlation: 0.81.

- [CATE Estimator Ranking on ACIC Setting A (p=0.1) 2016](<https://www.sota2.com/research/sota/cate-estimator-ranking-on-acic-setting-a-p-0-1-2016.md>): CATE Estimator Ranking; ACIC Setting A (p=0.1) 2016; Rank Correlation: 0.8.

- [CATE Estimator Ranking on ACIC Setting A (p=0.3) 2016](<https://www.sota2.com/research/sota/cate-estimator-ranking-on-acic-setting-a-p-0-3-2016.md>): CATE Estimator Ranking; ACIC Setting A (p=0.3) 2016; Rank Correlation: 0.8.

- [CATE Estimator Ranking on ACIC Setting B (xi=2) 2016](<https://www.sota2.com/research/sota/cate-estimator-ranking-on-acic-setting-b-xi-2-2016.md>): CATE Estimator Ranking; ACIC Setting B (xi=2) 2016; Rank Correlation: 0.77.

- [CATE Estimator Ranking on ACIC Setting C (m=0.1) 2016](<https://www.sota2.com/research/sota/cate-estimator-ranking-on-acic-setting-c-m-0-1-2016.md>): CATE Estimator Ranking; ACIC Setting C (m=0.1) 2016; Rank Correlation: 0.79.

- [CATE Estimator Ranking on ACIC Setting C (m=0.5) 2016](<https://www.sota2.com/research/sota/cate-estimator-ranking-on-acic-setting-c-m-0-5-2016.md>): CATE Estimator Ranking; ACIC Setting C (m=0.5) 2016; Rank Correlation: 0.81.

- [CATE Estimator Selection on ACIC Setting B, xi=2 2016](<https://www.sota2.com/research/sota/cate-estimator-selection-on-acic-setting-b-xi-2-2016.md>): CATE Estimator Selection; ACIC Setting B (xi=2) 2016; PEHE: 18.77.

- [CATE Estimator Ranking on ACIC Setting C (m=0.9) 2016](<https://www.sota2.com/research/sota/cate-estimator-ranking-on-acic-setting-c-m-0-9-2016.md>): CATE Estimator Ranking; ACIC Setting C (m=0.9) 2016; Rank Correlation: 0.8.

- [CATE Estimator Ranking on ACIC Setting B (xi=0) 2016](<https://www.sota2.com/research/sota/cate-estimator-ranking-on-acic-setting-b-xi-0-2016.md>): CATE Estimator Ranking; ACIC Setting B (xi=0) 2016; Rank Correlation: 0.85.

- [CATE Estimator Selection on ACIC Setting A, p=0 2016 (test)](<https://www.sota2.com/research/sota/cate-estimator-selection-on-acic-setting-a-p-0-2016-test.md>): CATE Estimator Selection; ACIC Setting A, p=0 2016 (test); Regret: not recorded.

- [CATE Estimator Selection on ACIC Setting A, p=0.1 2016 (test)](<https://www.sota2.com/research/sota/cate-estimator-selection-on-acic-setting-a-p-0-1-2016-test.md>): CATE Estimator Selection; ACIC Setting A, p=0.1 2016 (test); Regret: not recorded.

- [CATE Estimator Selection on ACIC Setting A, p=0.3 2016 (test)](<https://www.sota2.com/research/sota/cate-estimator-selection-on-acic-setting-a-p-0-3-2016-test.md>): CATE Estimator Selection; ACIC Setting A, p=0.3 2016 (test); Regret: not recorded.

- [CATE Estimator Selection on ACIC Setting B, xi=0 2016 (test)](<https://www.sota2.com/research/sota/cate-estimator-selection-on-acic-setting-b-xi-0-2016-test.md>): CATE Estimator Selection; ACIC Setting B, xi=0 2016 (test); Regret: not recorded.

- [CATE Estimator Selection on ACIC Setting B, xi=2 2016 (test)](<https://www.sota2.com/research/sota/cate-estimator-selection-on-acic-setting-b-xi-2-2016-test.md>): CATE Estimator Selection; ACIC Setting B, xi=2 2016 (test); Regret: not recorded.

- [CATE Estimator Selection on ACIC Setting C, m=0.1 2016 (test)](<https://www.sota2.com/research/sota/cate-estimator-selection-on-acic-setting-c-m-0-1-2016-test.md>): CATE Estimator Selection; ACIC Setting C, m=0.1 2016 (test); Regret: not recorded.

- [CATE Estimator Selection on ACIC Setting C, m=0.5 2016 (test)](<https://www.sota2.com/research/sota/cate-estimator-selection-on-acic-setting-c-m-0-5-2016-test.md>): CATE Estimator Selection; ACIC Setting C, m=0.5 2016 (test); Regret: not recorded.

- [CATE Estimator Selection on ACIC Setting C, m=0.9 2016 (test)](<https://www.sota2.com/research/sota/cate-estimator-selection-on-acic-setting-c-m-0-9-2016-test.md>): CATE Estimator Selection; ACIC Setting C, m=0.9 2016 (test); Regret: not recorded.

## Original sources and implementation

[Original paper](<https://arxiv.org/abs/2402.18392>)

[PDF](<https://arxiv.org/pdf/2402.18392.pdf>)
