Figure 6
From: Adjusting for sampling variability in sparse data: geostatistical approaches to disease mapping
Maps and scatterplots of the observed, Poisson kriging estimated, and UMBME estimated rates under the uniform assumption. Compared with the A) observed rate map, the B) Poisson kriging map displayed greater smoothing than the C) UMBME map (cases per 1,000 persons). However, scatterplots of the D) observed, E) Poisson kriging estimated, and F) UMBME estimated rates versus the true latent rate, X( s ), combined with MSE and LCCC calculations (Table 1), demonstrated that UMBME produced the highest estimation accuracy under the uniform assumption.