Quantitative assessment of probabilistic forecasts with applications in epidemiology

 

One of the main purposes of Bayesian hierarchical modelling is to make forecasts for future observations. In epidemiological applications, prediction problems are numerous. For example, forecasting future cancer rates based on cancer registry data is a central problem. In this paper I will describe methods for model choice and model criticism based on probabilistic predictions of external data. I do not intend to give a comprehensive overview over the literature on this subject, but will discuss some specific topics, in particular tools for multivariate forecasts. I will illustrate the methods through a case study from chronic disease epidemiology.