Dynamic Bayesian models for projecting cancer incidence in Puerto Rico
Pericchi Guerra, Luis Raúl
MetadataShow full item record
We estimate the present (2010) and predict the future (2014) of incidence for the top cancer tumor types in Puerto Rico (PR), by gender, age group and primary cancer site, to design public policy. Incidence data from Puerto Rico Central Cancer Registry were obtained for the years 1985 to 2004. The dynamic autoregressive models used in modern epidemiology are function of age-period-cohort (APC). We introduce a novel robust and stable prior the autoregressive variance, the scaled beta prior of the second kind (Beta2 prior). We found that this leads to a stable convergence of the model at the Markov Chain Monte Carlo (MCMC) implementation. We also produce statistical tools to check the goodness of fit and model selection.