Figure 8From: Adjusting for sampling variability in sparse data: geostatistical approaches to disease mappingMaps of the North Carolina HIV A) observed, B) Poisson kriging estimated, and C) UMBME estimated rates. While the Poisson kriging and UMBME maps displayed nearly identical spatial distributions of non-zero values, Poisson kriging exhibited greater smoothing strength than UMBME, thereby providing less distinction between areas of high and low rates (cases per 1,000 tests).Back to article page