Date of Award


Document Type

Open Access Thesis


Epidemiology and Biostatistics

First Advisor

Andrew Lawson


The effects of scale on the analysis of spatial data, often referred to as the modifiable areal unit problem in spatial studies, is one of the issues often encountered in small area health models. These spatial effects of scale are also seen in the areas of disease mapping where data are usually available in counts. Often there is a need to consider the different scales of aggregation that exist within count data, since inferences based on analyses can vary if we change the definition of the unit of analysis. This thesis provides a framework that describes the distribution of relative risk across a hierarchy of multiple scales. With the help of simulation studies, we explore a methodology that allows us to estimate and compare measures of relative risk in Poisson-based models for count data. The proposed method will be illustrated through the Georgia Oral Cancer data set (2004), which has count data at two levels: County and Public health district.