Some problems in geographical epidemiology
The analysis of the geographical variation of disease has been part of the subject of epidemiology since its beginnings. However, it has not been the subject of much statistical research until the late 1980s when it became computationally feasible to examine the spatial structure. In thesis three problems of geographical epidemiology are examined, namely: 1. A simulation study which examines how different models used for disease mapping perform under various 'true' relative risk surfaces. 2. A new model for disease mapping is proposed which is based on the spatial variation being split into two different types of autocorrelation. These different types of spatial autocorrelation can be used in order not to over smooth any rural/urban divide which might exist in the dataset. 3. A new model is proposed for the examination of clustering in the spatio-temporal domain. Although a number of different tests exist for the detection of clustering in spatio-temporal domain, to date, very little has been done on the modelling of the clustering.