Simultaneous estimation and model choice for big discrete time-to-event data with additive predictors

Benjamin Müller, University of Innsbruck

Co-authors: Nikolaus Umlauf, University of Innsbruck; Johannes Seiler, University of Innsbruck; Kenneth Harttgen, ETH Zurich; Stefan Lang, University of Innsbruck

Abstract: This work introduces an extension of the novel batchwise backfitting algorithm for the estimation of additive discrete time-to-event models. The algorithm is designed to efficiently handle large datasets while incorporating a boosting-type approach for automated variable selection. An extensive simulation study is conducted to evaluate the performance of the algorithm against established methods. The effectiveness is further demonstrated by modelling infant mortality in ten Eastern sub-Saharan African countries. The results highlight the algorithm’s strong estimation performance, excellent variable selection capabilities and scalability for large datasets.