ABSTRACT
Disease mapping has a long history in epidemiology, starting with John Snow's map of cholera cases in London in 1854. In terms of the areas used for disease mapping, there will be a trade-off in relation to the geographical scale that is used: rates calculated using larger geographical areas will be more stable but summaries of risk, such as relative risks, may be affected by the aggregation of large numbers of individuals. For small areas, and in particular rare diseases, estimates of risk may be dominated by sampling variability. The Poisson-gamma model is analytically tractable, but does not easily allow the incorporation of spatial random effects.
