Risk-Calibrated Balancing for High-Dimensional Causal Extrapolation
Abstract
In observational causal inference, covariate balancing is widely used to reduce source-target covariate shift, but under weak overlap in high dimensions, stronger balance can induce concentrated weights and increase variance. Balance measures how well the target covariate distribution is represented, but does not by itself determine how reliably the counterfactual mean can be estimated. We develop risk-calibrated balancing for the average treatment effect on the treated, which applies ridge augmentation to any normalised base weights and selects its penalty using conditional prediction risk of the counterfactual mean. Under a random-effects predictive model, we derive an exact finite-sample decomposition of this risk into residual covariate imbalance and weight-induced variance. For design-independent base weights under proportional asymptotics, we characterise how limiting risk depends on source and target covariance geometry, population mean shift, and weight concentration. For covariate-adaptive base weights, we develop a uniformly consistent target-aware risk estimator whose minimiser attains vanishing scaled oracle excess risk. Simulations show that the high-dimensional risk predictions remain informative for adaptive balancing and that target-aware tuning generally reduces excess target risk. Empirical analyses of job-training and single-cell perturbation data show that risk-calibrated balancing generally improves on the corresponding base estimators, with larger gains under weaker overlap.
Related
- Proximal Balancing for Causal Effect Estimation under Unmeasured Confounding
- Calibrated Bayesian inference for overall and subgroup-specific causal effects in stepped-wedge cluster-randomized trials
- Proximal causal inference through cross-proxy balancing
- Into the danger zone: stable extrapolation in high-dimensional function and operator learning
- Analytically Calibrated Upper L-Statistics for High-Dimensional Mean Change-Point Testing
BibTeX
@misc{yang2026riskcalibrated,
title = {{Risk-Calibrated Balancing for High-Dimensional Causal Extrapolation}},
author = {Fenglin Yang and Haoran Lei and Yan Chen and Jin-Hong Du},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2609.34482},
eprint = {2609.34482},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2609.34482},
}
RIS
TY - UNPB
TI - Risk-Calibrated Balancing for High-Dimensional Causal Extrapolation
AU - Fenglin Yang
AU - Haoran Lei
AU - Yan Chen
AU - Jin-Hong Du
PY - 2026
DA - 2026/10/02
PB - arXiv
DO - 10.48550/arxiv.2609.34482
UR - https://arxiv.org/abs/2609.34482
AB - In observational causal inference, covariate balancing is widely used to reduce source-target covariate shift, but under weak overlap in high dimensions, stronger balance can induce concentrated weights and increase variance. Balance measures how well the target covariate distribution is represented, but does not by itself determine how reliably the counterfactual mean can be estimated. We develop risk-calibrated balancing for the average treatment effect on the treated, which applies ridge augmentation to any normalised base weights and selects its penalty using conditional prediction risk of the counterfactual mean. Under a random-effects predictive model, we derive an exact finite-sample decomposition of this risk into residual covariate imbalance and weight-induced variance. For design-independent base weights under proportional asymptotics, we characterise how limiting risk depends on source and target covariance geometry, population mean shift, and weight concentration. For covariate-adaptive base weights, we develop a uniformly consistent target-aware risk estimator whose minimiser attains vanishing scaled oracle excess risk. Simulations show that the high-dimensional risk predictions remain informative for adaptive balancing and that target-aware tuning generally reduces excess target risk. Empirical analyses of job-training and single-cell perturbation data show that risk-calibrated balancing generally improves on the corresponding base estimators, with larger gains under weaker overlap.
ER -