Figure 2.

Impact of population size and local variance on the shrinkage factor in empirical Bayes smoothing. The local variance is the population-weighted variance among the 32 closest breast cancer mortality rates for each county. The shrinkage factor represents the relative weight assigned to the local rate versus the local mean in the local empirical Bayes smoother. This factor increases with the population size of the county (i.e. more reliable rate) and in regions where the local variance is high. Selected counties are depicted by orange dots in the scattergram and orange boundaries on the map of shrinkage factor.