Abstract:
|
Learning optimal individualized treatment regime has recently attracted a lot of attention for personalized medicine. In this talk, I will present a doubly robust estimation method for estimating optimal individualized treatment regime in additive hazards regression with censored survival data. By properly adjusting for time-dependent propensity scores, the new method is robust against the misspecification of the main effects of covariates, and therefore enjoys the doubly robust property as in the A-learning estimation for uncensored data. Simulations and a real data application will be shown to illustrate the performance of the proposed method.
|