How does ArcGIS desktop estimate prediction values (out-of-sample) using a GWR calibration? Does it use the closest calibrated point? Does it use raster surfaces build with coefficients? Other Method?
Not sure if the answer is here (I'm not a statistician), but take a peek if you have not already:
Geographically Weighted Regression (GWR)—Help | ArcGIS for Desktop
Adding Spatial Statistics for greater exposure.
Chris Donohue, GISP
Thanks for the information Chris. Unfortunately, there is nothing in the ArcGIS Help for GWR outlining the method to predict values for out-of-sample points which were not used in the initial calibration of GWR.
I will post any information any information I find.
Retrieving data ...