ABSTRACT

The spatial analysis of single diseases is often sufficient. However, in some applications there is a need to consider different scales of aggregation within an analysis.

One such situation arises when it is of interest to consider a relationship at different aggregation levels. For example, if the relation of an outcome at county level to a covariate is examined, will the relationship hold true at lower aggregation levels (e.g., census tract) or at higher levels (e.g., state or country)? In general, it is unlikely that this would be the case as, if it were, there would be little need to consider different levels of analysis. In fact ecological bias would not occur. Scale change issues are often known as the modifiable areal unit problem (MAUP), whereby modification of the areal units could lead to different inferences. In geostatistics, this is known as the change of support problem (Cressie, 1996; Banerjee et al., 2004).