Spatial modelling of the relative risk of dengue fever in Rio de Janeiro for the epidemic period between 2001 and 2002

We model the number of cases of dengue fever in Rio de Janeiro, Brazil, between the 48 epidemiological week of 2001 and the 20 epidemiological week of 2002. This period comprises the worst dengue fever epidemic the city has ever experienced. The original data were aggregated across time and classified according to the district where the infected person lived. We consider the number of cases of the disease for each district as following a Poisson random variable. We compare among many different models that consider social-economical and infrastructural covariates plus random effects to explain the logarithm of the mean of this Poisson. Following the Bayesian paradigm, we consider these random effects to follow a Conditional Autoregressive (CAR) prior. Our main contribution resides on entertaining different neighborhood structures for these effects. These different neighborhood structures take into account the particular landscape of the city of Rio de Janeiro, which is known for its high mountains. We have observed some significant effect of social-economics and infra structural factors on the relative risk of the disease. The great variety of results that we provide has given evidence that the choice of the neighborhood structure when the CAR prior is used, might be of importance, as the significance of the covariates changes, when we use different neighborhood structures.