Covariate-Balancing-Aware Interpretable Deep Learning models for Treatment Effect Estimation
This work addresses the need for interpretable and accurate treatment effect estimation in biomedical applications, representing an incremental improvement over existing methods.
The authors tackled the problem of estimating average treatment effects from observational data by proposing a novel objective function based on energy distance balancing scores, which avoids the need for correct propensity score specification and improves interpretability using neural additive models. Their model demonstrated superior performance over state-of-the-art methods in semi-synthetic experiments on benchmark datasets IHDP and ACIC, with tighter theoretical bias bounds reported.
Estimating treatment effects is of great importance for many biomedical applications with observational data. Particularly, interpretability of the treatment effects is preferable for many biomedical researchers. In this paper, we first provide a theoretical analysis and derive an upper bound for the bias of average treatment effect (ATE) estimation under the strong ignorability assumption. Derived by leveraging appealing properties of the Weighted Energy Distance, our upper bound is tighter than what has been reported in the literature. Motivated by the theoretical analysis, we propose a novel objective function for estimating the ATE that uses the energy distance balancing score and hence does not require correct specification of the propensity score model. We also leverage recently developed neural additive models to improve interpretability of deep learning models used for potential outcome prediction. We further enhance our proposed model with an energy distance balancing score weighted regularization. The superiority of our proposed model over current state-of-the-art methods is demonstrated in semi-synthetic experiments using two benchmark datasets, namely, IHDP and ACIC.