A0932
Title: Estimating optimal treatment regimes in semi-supervised framework
Authors: Mengjiao Peng - East China Normal University, China (China) [presenting]
Abstract: Finding the optimal individualized treatment rule has been studied intensively in the literature, with important applications in practice. The problem of estimating the optimal treatment regime in a semi-supervised learning setting is considered, where a very small proportion of the entire set of observations is labelled with the true outcome but features predictive of the outcome are available among all observations. A model-free robust inference approach for optimal treatment regimes is proposed with the aid of the unlabeled data with only covariate information to improve estimation efficiency. The proposed estimation of OPT primarily involves a flexible nonparametric imputation by single index kernel smoothing which works well even for high-dimensional covariates, and a follow-up estimation for optimal treatment regime based on concordance-assisted learning, including optimization of the estimated concordance function up to a threshold and finding the optimal threshold to maximize the inverse propensity score weighted (IPSW) estimator of the value function. Moreover, when the propensity score function is unknown, a doubly robust estimation method is developed under a class of monotonic index models. The estimators are shown to be consistent and asymptotically normal. Simulations exhibit the efficiency and robustness of the proposed method compared to existing approaches in finite samples.