Hi members, I'd like to ask a question. I've run some regression models (OLS, ML) with Geoda program. I found high multicollinearity with one (!) independent variable. (Multicollinearity Condition Number > 30.) How could this be? What's the explanation for this phenomenon? Thank you very much!
Hello Zoltan,
The reason is that the GeoDa program implements spatial regression, not ordinary least squares regression. The spatial lag/proximity is used so as to remove possible autocorrelation from the residuals. If you ignored the spatial information, and ran an ordinary OLS regression, you wouldn't have that result (but, of course, there could be autocorrelation among the residuals, which violates one of the OLS assumptions).
See this page for an explanation: http://www.bisolutions.us/A-Brief-Introduction-to-Spatial-Regression.php
Good luck with your work.
- Badges
- Science topic
- Similar topics
- Papers