Spatial regression

Abstract

The estimation of risk indices at particular geographic locations plays an important role in epidemiological investigation of injury. Until the advent of geocoded injury data and computerized geographic information systems, the majority of spatial studies concentrated on describing the distribution of injury morbidity and mortality across a region by manually plotting event data on a map. Little progress was made in developing the tools and methodologies devoted to studying the spatial interactions between agent, host, and environmental contributing factors. Geographically weighted regression (GWR) was recently introduced as a method for analyzing the relationship between two or more variables that vary over geographic space. This chapter introduces the fundamental concepts of GWR by contrasting it to linear regression as well as presenting an example of its application to emergency department utilization data. While the GWR method has been underutilized in the study of injury, increased use of this technique in the future will undoubtedly add a unique perspective to the field of injury research.