An analysis of Japanese liver cancer mortality data with Bayesian age–period–cohort models

Age–period–cohort (APC) models have been widely used in the analysis of incidence and mortality data. Bayesian APC models, in which multivariate Gaussian priors are incorporated on age, period and cohort effects, can evade the identifiability problem. Inference with integrated nested Laplace approximations (INLA) has recently been a useful tool. An application of the Bayesian APC models with INLA to Japanese liver cancer mortality data is illustrated, in which a sudden change of the cohort effect was revealed. Keyword: information criteria; integrated nested Laplace approximation; Gaussian Markov random field