CMStatistics 2022: Start Registration
View Submission - CMStatistics
B0248
Title: Bootstrapping complex survival models: From type-II censoring to causal inference Authors:  Sarah Friedrich - University of Augsburg (Germany) [presenting]
Jasmin Ruehl - University of Augsburg (Germany)
Abstract: Bootstrap methods are frequently applied to derive confidence intervals in complex survival settings. The classical nonparametric bootstrap by Efron, however, relies on the independence assumption, which is not always fulfilled. For example, randomised clinical trials with time-to-event endpoints are frequently stopped after a pre-specified number of events has been observed. This practice leads to dependent data and non-random censoring, though, which can generally not be solved by conditioning on the underlying baseline information. Matters are further complicated by staggered study entry. Our simulations show that a martingale-based wild bootstrap approach still provides reasonable estimates, while Efron's classical bootstrap may lead to biased results. In the context of causal effect estimates in competing risks settings, bootstrap approaches are also often applied. We investigate the asymptotic validity of the bootstrap in these settings and compare different approaches by means of simulations.