Authors: Debbie Christian*, University of North Texas
Topics: Geographic Information Science and Systems, Medical and Health Geography, Applied Geography
Keywords: GIS, data aggregation, geospatial data, kernal denstiy estimation
Session Type: Paper
Start / End Time: 1:20 PM / 3:00 PM
Room: Rampart, Sheraton, 5th Floor
Presentation File: No File Uploaded
Disease maps depict the spatial variations in disease risk and can be produced in a number of ways. Choropleth maps of disease outcomes generally portray rates using predefined areal units and shades of color to represent intensity. However, choropleth maps are not desirable because the patterns observed are influenced by the choice of spatial unit. This is well-known as the modifiable areal unit problem (MAUP). Further, areal units with small populations are known to yield unstable estimates of disease rates due to the small numbers problem. These issues are generally addressed through the use of smoothing methods including spatial filtering (fixed or adaptive), Bayesian techniques, or geostatistical approaches. Maps produced using adaptive spatial filters are designed to maintain approximately equal population sizes for each rate estimate on the map, thus ensuring a minimum level of statistical reliability. However, if the underlying population data are aggregated to some small area unit such as a zip code, resulting maps may continue to portray different levels of statistical reliability. This is due to the implementation of the adaptive filters method that is designed to guarantee a minimum lower bound on population size with no limits on the upper bound. In this presentation, I present a modification to the adaptive spatial filter algorithm that allows the user to develop maps with a more uniform levels of statistical reliability by providing finer controls on the population sizes used to calculate each rate estimate.