Figure 1
From: Relevance of the type III error in epidemiological maps
Austrian infant mortality from 1984 to 2008 at the level of 121 administrative districts. SMRs have been estimated by the empirical Bayes procedure and have been grouped into quartiles. The expression "true SMR" refers to the use of these results as calibration reference points for type III error calculations in the example. Kernel estimates of the distributions of SMRs and expected cases are also shown.