ABSTRACT
The natural boundaries of a disease which encompass areas with suitable conditions for its transmission are often unknown. In this chapter, the authors introduce a general modelling framework that deals with zero-inflation and discuss several main forms of spatially structured zero-inflation for disease counts data. They describe methods of inference and spatial prediction and illustrates such applications in river-blindness and Loa loa mapping. For the sake of simplicity and without loss of generality, the authors describe different forms of zero-inflation by restricting their attention to disease prevalence data. From a statistical perspective, the zero-inflated model is an improvement over the standard model because its additional flexibility allows for more complex patterns in disease prevalence. The results of this analysis suggest that data show indeed a higher level of over-dispersion than that implied from the standard geostatsitical model for prevalence.
