Figure 9.
Impact of ignoring the spatial support and population distribution on Poisson kriging results. Top scatterplots illustrate the similarity of the cancer mortality risk estimated using punctual and area-to-area (ATA) Poisson kriging. Ignoring the shape and size of counties (i.e. collapsing county rates to their population-weighted centroids) leads to an overestimation of the kriging variance, in particular for the set of vastly different counties in Region 2.