Sources of Gain: Decomposing Performance in Conditional Average Dose Response Estimation

TL;DR

This paper introduces a decomposition scheme to evaluate the performance of CADR estimators, revealing that common benchmarks often do not accurately reflect the challenges estimators aim to address, especially confounding.

Contribution

It proposes a novel method to disentangle different factors affecting CADR estimator performance and applies it to evaluate popular estimators on benchmark datasets.

Findings

Most benchmarks do not primarily test for confounding.

Confounding is not a significant issue in the evaluated datasets.

Benchmark challenges often differ from their intended purpose.

Abstract

Estimating conditional average dose responses (CADR) is an important but challenging problem. Estimators must correctly model the potentially complex relationships between covariates, interventions, doses, and outcomes. In recent years, the machine learning community has shown great interest in developing tailored CADR estimators that target specific challenges. Their performance is typically evaluated against other methods on (semi-) synthetic benchmark datasets. Our paper analyses this practice and shows that using popular benchmark datasets without further analysis is insufficient to judge model performance. Established benchmarks entail multiple challenges, whose impacts must be disentangled. Therefore, we propose a novel decomposition scheme that allows the evaluation of the impact of five distinct components contributing to CADR estimator performance. We apply this scheme to eight popular CADR estimators on four widely-used benchmark datasets, running nearly 1,500 individual experiments. Our results reveal that most established benchmarks are challenging for reasons different from their creators' claims. Notably, confounding, the key challenge tackled by most estimators, is not an issue in any of the considered datasets. We discuss the major implications of our findings and present directions for future research.