[Openspace] test on remaining autocorrelation
koschins at uiuc.edu
Fri Apr 28 14:31:08 CDT 2006
Yes, Geoda does not contain the tests for remaining spatial autocorrelation
yet - this is one of those things on our "to do" list ...
The likelihood ratio test is comparable to the LM test in that it assesses
whether spatial autocorrelation is a problem in your non-spatial/OLS model.
The reason why it gets listed after the spatial model output is because it
requires the estimation of the spatial model's log likelihood (which is
compared to that of the null/the OLS model). Since the LM test is about the
difference between slopes of the likelihood function (which is zero for the
spatial model), you only need the OLS model to compute the LM test, so the
LM results are accessible without estimating a spatial model.
From: openspace-bounces at sal.uiuc.edu [mailto:openspace-bounces at sal.uiuc.edu]
On Behalf Of V.A. Ceccato
Sent: Friday, April 28, 2006 5:30 AM
To: openspace at sal.uiuc.edu
Subject: [Openspace] test on remaining autocorrelation
How do you test for remaining autocorrelation in the Lag model?
My spatial diagnostics of OLS model showed clear indication of
autocorrelation on residuals, so I apllied the lag model to try to get rid
of this problem. However, regardless what I do to the model, The Likelihood
Ratio Test comes always up as significant. In the GeoDa Workbook, page 209,
it says "The (second) test is an alternative to the asymptotic significance
test on the spatial autoregressive coefficient, it is NOT a test on
remaining spatial autocorrelation."
So, how do we know whether or not the problem of autocorrelation was solved?
Vania Ceccato Ps. The test on remaining autocorrelation was available in
SpaceStat, why is it gone in GeoDa?
Openspace mailing list
Openspace at sal.uiuc.edu
More information about the Openspace