ABSTRACT
This chapter reviews some basic disease mapping models and their implementation. When a bivariate realization of cases and controls are available it is possible to make conditional inference on this joint realization. This leads to a simpler conditional logistic model formulation. A special case of the binomial model arises when the disease outcome is binary and so the binomial model simplifies to a Bernoulli model. In this case individual level predictors could be available but the geocoding could be contextual in that it may be the case that only an aggregated spatial unit can be used. Perhaps the most commonly encountered model for small area count data is the Poisson model. This model is appropriate when there is a relatively low count of disease and the population is relatively large in each small area. The expected rates or counts that are used in Poisson models are usually assumed to be fixed quantities. They are a product of indirect standardization.
