Impact of population size and local variance on LBS and PK results. The top scattergrams illustrate the increasing weight assigned to the original mortality rate (i.e. shrinkage factor or kernel weight) as the population in that county increases. The local empirical Bayes (LBS) and Poisson kriging (PK) weights are more distinct in zones of greater variability in mortality rates, as measured by the local variance. Differences between kernel weights are not correlated with the differences between the corresponding risk estimates.