# Problems with spatial autocorrelation in Geographically Weighted Regression (GWR) analysis [closed]

I have conducted an Ordinary Least Squares (OLS) and Geographically Weighted Regression (GWR) analysis in ArcGIS where I have tried to predict biodiversity patterns in an area. When I examine the results with Spatial Autocorrelation the residuals tends to be clustered with a fairly high Z-value.

The result from the GWR, however, tends to be the expected when I visualize it. How do I handle the clustering? What problems does it creates, IS there any problems?

## closed as too broad by ahmadhanb, BBG_GIS, BERA, aldo_tapia, MaryBethNov 13 '17 at 13:02

Please edit the question to limit it to a specific problem with enough detail to identify an adequate answer. Avoid asking multiple distinct questions at once. See the How to Ask page for help clarifying this question. If this question can be reworded to fit the rules in the help center, please edit the question.

• Could you please clarify what it means to "examine the results with spatial autocorrelation" and how that indicates clustering of the residuals? Showing a small example might help. – whuber Aug 25 '13 at 20:17
• Well, I followed the Esri tutorial on GWR and it seems to be recommended to check the result from the std residuals with spatial autocorrelation in order to see if they tend to be clustered, random or dispersed. – user21070 Aug 26 '13 at 21:54
• So what you're saying is that you have computed a measure of clustering of the GWR residuals and it appears to indicate clustering. What happens when you decrease the size of the local neighborhood used in GWR? Does the clustering appear to be reduced? – whuber Aug 26 '13 at 23:00
• Well, yes. If I decrease the size of the local neighborhood, the clustering appears to be reduced, that´s correct. The problem is, however, that when I decrease the number of neighbours the GWR doesn´t seem to produce a full-scale analysis and it produces blanks among the polygon. Um, do you understand? – user21070 Aug 27 '13 at 7:26
• Yes, that is to be expected. It indicates you might not have sufficient data density in order to carry out GWR. It's not possible to recommend alternatives without knowing why you're performing GWR, what you hope to achieve with it, and the nature of your data. – whuber Aug 27 '13 at 15:11