Bayesian disease-mapping models treat the relative risks {θi } as random variables and specify a distribution for them. This part of the model is crucial, as the distributional assumptions thus made allow borrowing of information across the areas. The distribution specified is referred to as the second hierarchical level of the model to distinguish it from the first-level distribution specified in equation 1 that pertains to the random sampling variability of the observed counts about their local mean. It is at this second level that the spatial dependence between the relative risks is introduced. This spatial dependence is represented by means of a prescribed neighborhood graph that defines the set of neighbors (denoted by ∂i) for each area i.