Title: Copula link-based additive models for dependent right-censored event time data
Authors: Giampiero Marra - University College London (United Kingdom)
Rosalba Radice - Cass Business School (United Kingdom)
Robinson Dettoni - University College London (United Kingdom) [presenting]
Abstract: When time to event data is analysed is often assumed that the censoring mechanism is independent. This can be appropriate in many situations, in particular, when individuals are censored at the end of the study. However, in many applications, this assumption can be challenged. The aim is to introduce a class of flexible survival models in which the censoring scheme is dependent, non-informative and there are not competing risks. In particular, we show that our model is identified. Baseline functions for the event and censored times are non-parametrically estimated using monotonic P-splines. In addition, covariate effects are flexibly determined using additive predictors that allow for a vast variety of covariate effects, whereas parameter estimation is reliably carried out within a penalised maximum likelihood framework with integrated automatic multiple smoothing parameter selection. The square root (n)-consistency and asymptotic normality of the proposed flexible dependent estimator are derived. The finite sample properties of the estimators are investigated via a Monte Carlo simulation study which highlights the bias when dependent censoring is ignored and the good empirical performance of our framework. The proposal is illustrated using liver transplants data. The discussed models and methods have been implemented in the R package GJRM to allow for transparent and reproducible research.