Modelling mortality data on the Lexis diagram

Currie, Durban & Eilers (2004) used 2-dimensional P -splines to smooth mortality data classified by age at death and year of death. In this paper we apply this model to data classified by age at death, year of death and year of birth. Discrete cohort effects are added to the model using a method similar to the overdispersion model of Perperoglou and Eilers (2006). This model allows us to decompose the mortality surface into (i) a smooth 2-dimensional surface in age and time and (ii) discrete cohort effects. We illustrate our remarks with the analysis of some German mortality data.