A1110
Title: On multiple robustness of proximal dynamic treatment regimes
Authors: Yuanshan Gao - Center for Data Science, Zhejiang University (China) [presenting]
Yang Bai - National University of Singapore (Singapore)
Yifan Cui - Zhejiang University (China)
Abstract: Dynamic treatment regimes (DTRs) are sequential decision rules that adapt treatment according to individual time-varying characteristics and outcomes to achieve optimal effects, with applications in precision medicine, personalized recommendations, and dynamic marketing. Estimating optimal DTRs via sequential randomized trials might face costly and ethical hurdles, often necessitating the use of historical observational data. The purpose is to utilize a proximal causal inference framework for the identification of optimal DTRs when the unconfoundedness assumption fails. The contributions are three-fold: (i) proposing a new non-parametric identification method for optimal DTRs with a reduced risk of error amplification; (ii) establishing the semi-parametric theory, including efficient bounds, for the value function of a given sequence of treatment rules; and (iii) proposing a multiply robust method for identifying and estimating optimal DTRs and proving the corresponding theoretical properties such as consistency and convergence rate. Numerical experiments validate the efficiency and multiple robustness of our proposed methods.
