Mapping disease risk estimates based on small numbers: an assessment of empirical Bayes techniques
Item Type:Journal Article
Citation:Pringle, D.G.. 'Mapping disease risk estimates based on small numbers: an assessment of empirical Bayes techniques'. - Economic & Social Review, Vol. 27, No. 4, July, 1996, pp. 341-363, Dublin: Economic & Social Research Institute
27 jul 96 pringle.pdf (Published (publisher's copy) - Peer Reviewed) 1.211Mb
Choropleth maps are frequently used to analyse spatial variations in the risk of a disease. In such maps the relative risk is typically quantified by dividing some measure of the number of cases of the disease by some measure of the population at risk. The resulting rates may be regarded as maximum likelihood estimates of individual risk. These estimates may be unstable if the areas are very small or if the disease is rare. I n such situations, the highest and lowest values on the map will display a tendency to be concentrated in the areas with the smallest populations. The traditional solution to this problem is to supplement maps based on ratios with probability maps. However, probability maps display an opposite bias ? i.e., they tend to highlight the areas with the largest populations. Several statisticians have suggested a compromise between these two extremes using empirical Bayes techniques. This paper outlines the rationale underlying empirical Bayes techniques, and assesses their usefulness using case studies of neo-natal mortality and cancer mortality.
Author: Pringle, D. G.
Publisher:Economic & Social Research Institute
Type of material:Journal Article
Availability:Full text available