rosekell

non-linearity of data

Discussion created by rosekell on Feb 18, 2013
Hi all you experts in GWR out there!
I am currently modeling urban land use change- trying both a spatial lag and spatial autoregressive model- in Stata. I would like to compare the results of such models using the new GWR capabilities in ArcGIS 10.2, but first I have to address this problem: I am faced with a suite of explanatory variables that I know should be, and are in theory, highly significant predictors of urban land use change (including: share of forest, agriculture, and other remote sensing data (from corine land cover, like Landsat), soil quality, slope and elevation indicators, and share of protected areas. Unfortunately, all of these display non linear relationships with the response variable of urban land use change. So. Is there a non-linear model of geographically weighted regression? Or, is transforming my data into logged variables the best solution to try and fit the relations to a "line."

Thanks so much!

Outcomes