Copula link-based additive models for bivariate time-to-event Outcomes with general censoring scheme: Computational advances and variable ranking procedures

Abstract

Bivariate survival outcomes frequently arise in applied studies where the occurrence of two associated events of interest is observed. However, the practical utility of bivariate copula survival models is often hindered by the presence of data influenced by various censoring mechanisms and high-dimensional datasets. This thesis presents a novel solution to address these two challenges. In the first part, we propose a general and flexible copula regression approach capable of effectively handling bivariate survival data subject to various censoring mechanisms. This approach offers versatility in modelling the association between the two events of interest. In the second part, we introduce a variable selection procedure based on the class of models presented in the previous section. This procedure represents an absolute novelty in the panorama of bivariate copula survival models. To evaluate the effectiveness of the proposed methodologies, extensive simulation studies are conducted. Additionally, illustrated examples using data from the Age-Related Eye Disease Study are discussed. Lastly, the developed modelling frameworks are implemented in the R programming language, making them accessible to a wide range of users. [edited by Author]