Title: Wild bootstrap confidence bands for the cumulative incidence function in Fine-Gray models
Authors: Marina Tiana Dietrich - Vrije Universiteit Amsterdam (Netherlands) [presenting]
Dennis Dobler - Vrije Universiteit Amsterdam (Netherlands)
Mathisca de Gunst - Vrije Universiteit Amsterdam (Netherlands)
Abstract: The Fine and Gray model in the competing risks setting with censoring-complete data is considered. The goal is to construct asymptotically valid time-simultaneous confidence bands for the cumulative incidence function using the wild bootstrap. To verify the underlying theory, they use elegant and general martingale arguments. Especially for small samples, the flexibility of the wild bootstrap procedure with possibly non-normal multipliers seems preferable over a Gaussian approximation because the latter is only aiming at the asymptotic distribution. The performance of the wild bootstrap confidence bands for different types of multipliers and weight functions is empirically studied through simulations. In addition, they illustrate the developed method by investigating the impact of Pneumonia for intensive care unit patients on the probabilities of alive discharge competing with hospital death.