Implementing Exploratory Spatial Data Analysis Methods For..

This paper reports on the development of prototype software designed for exploratory visualization of geographically referenced health statistics. The software prototype provides a number of interactive methods for exploring relationships between risk factors and mortality rates and how they are distributed in space. The use of geographically referenced mortality data to detect disease "hot spots" can be traced, at least, to Dr. John Snow’s 1854 map of cholera deaths in London, which allowed him to hypothesize that a particular water pump was the source of the epidemic. While the use of traditional static maps for cluster identification continues to be important, with a major new atlas of mortality in the U.S. just published, dynamic exploratory data analysis and visualization techniques have the potential to further enhance detection of "hot spots". Our prototype implements a number of exploratory data visualization techniques within existing geographic information systems software (ArcView GIS, ESRI, Redlands, CA). These techniques, including scatterplot brushing, interactive data classification, focusing, and representational methods for multivariate display, can help the analyst identify disease "hot spots" and facilitate data exploration that may lead to hypothesis about causal links between morality and potential risk factors. This paper will discuss the technical issues surrounding the implementation of these visualization techniques and the ways in which these techniques may be employed in existing GIS software. Finally, it will examine possible enhancements to the methods implemented here.