B1047
Title: Communication-efficient distributed estimation and inference for Cox's model
Authors: Zhipeng Lou - University of Pittsburgh (United States) [presenting]
Abstract: Motivated by multi-center biomedical studies that cannot share individual data due to privacy and ownership concerns, communication-efficient iterative distributed algorithms are developed for estimation and inference in the high-dimensional sparse Cox proportional hazards model. The estimator, with a relatively small number of iterations, is demonstrated to achieve the same convergence rate as the ideal full-sample estimator under very mild conditions. To construct confidence intervals for linear combinations of high-dimensional hazard regression coefficients, a novel debiased method is introduced, central limit theorems are established, and consistent variance estimators are provided that yield asymptotically valid distributed confidence intervals. In addition, valid and powerful distributed hypothesis tests are provided for any of its coordinate elements based on decorrelated score tests. Time-dependent covariates are allowed as well as censored survival times. Extensive numerical experiments on both simulated and real data lend further support to the theory and demonstrate that the communication-efficient distributed estimators, confidence intervals, and hypothesis tests improve upon alternative methods.
