A unifying framework for flexible excess hazard modelling with applications in cancer epidemiology

Excess hazard modelling is one of the main tools in population-based cancer survival research. Indeed, this setting allows for direct due to even absence reliable information on cause death, which common epidemiology studies. We propose a unifying link-based additive framework excess that inclusion many types covariate effects, including spatial and time-dependent using any type smoother, such as thin plate, cubic splines, tensor products Markov random fields. In addition, accounts all censoring well left truncation. Estimation conducted by an efficient stable penalized likelihood-based algorithm whose empirical performance evaluated through extensive simulation Some theoretical asymptotic results are discussed. Two case studies presented data from patients diagnosed with breast (female), colon lung cancers England. The support presence non-linear effects variation. proposed approach available R package GJRM.