Abstract

When event rates exhibit significant disparities between age groups, steps must be taken to ensure fair comparisons between geographic areas with disparate age distributions. One way to achieve this is to calculate age-specific estimates of the event rates in each area and then use the age distribution of a standard population (e.g., the 2010 US Standard Population) to weight the estimates and calculate the age-standardized estimates. When population sizes and/or death counts are small, these age-specific rate estimates may be unstable, thus it might be desirable to first model the age-specific death data to produce more stable estimates before conducting the age-standardization. Recent work, however, has revealed the extent to which commonly used spatial models can produce overly precise (and overly smooth) estimates, an issue that is compounded when analyzing a collection of age-specific datasets. In this talk, we will begin by illustrating how the effect of prior information in a Bayesian analysis of age-specific event data combines to influence the posterior distribution of the age-standardized estimates. We will then demonstrate how a multivariate conditional autoregressive (MCAR) model can be designed to account for dependencies between age groups while also controlling the informativeness of the model to prevent oversmoothing.