Geographically Weighted Regression (GWR) is a method of analysing spatially varying relationships. This usually involves fitting a model to predict the values of one variable (response or dependent variable) from a set of one or more independent (predictor) variables. The new Geographically Weighted Regression Tool in the Spatial Statistics Toolbox permits you to fit GWR models and display the results of the analysis using the features of ArcGIS 9.3.
The GWR Tool has been developed by ESRI (the Environmental Systems Research Institute of Redlands, CA) in close cooperation with the original developers of GWR, Professor A Stewart Fotheringham and Martin Charlton of the National Centre for Geocomputation (NUI Maynooth) and Professor Chris Brunsdon of the University of Leicester.
Charlton and Fotheringham have also authored some extra material to help users of GWR obtain the maximum from their analyses:
The White Paper describes the GWR approach in some detail, together with extensive information on the theoretical background to GWR. The Tutorial takes the user step by step through the processes of creating data for, fitting and calibrating, displaying and interpreting the results of a GWR analysis.
You can access the GWR tool if you have an ArcInfo license, or a Spatial Analyst license, or a Geostatistical Analyst license. If you have an ArcView or an ArcEditor license, you will need to obtain either Spatial Analyst or Geostatistical Analyst to run the GWR Tool.