Linking spatial data from different sources: the effects of change of support

A nationwide Environmental Public Health Tracking program is being created to monitor environmental impacts on human health. This, and many other efforts to relate environmental and health outcomes, depend largely on the synthesis of existing data sets; little new data are being generated for this purpose. More often than not, the data available for such synthesis have been collected for different geographic or spatial units, and any set of these units may be different from the one of interest. In this paper, we compare and contrast two approaches that can be used within a Geographic Information System to link spatial data from different sources. The first approach works with centroids of areal units and is commonly used in environmental health analyses. The second approach honors the spatial support (size, shape and orientation) of the data. Using traditional regression models and a spatially-varying coefficient regression model, we show that different linkage methods can lead to different inference. We describe key ideas pertaining to the support of spatial data that are often ignored in many analyses of environmental health data and present a general analytical approach to change-of-support problems.