GWR and Jarque-Bera Statistic

Discussion created by lkh344 on Jun 1, 2013
Hello All,

I ran OLS model before checking the variations of the relationship using GWR. The OLS model shows that Jarque-Bera value is not significant but BP value is significant. So I interpreted the OLS model need to check non-stationality by using GWR. However, there is no model biased problem.

The other model shows different statistical diagnostics (please see the below). My questions is are the following: if both BP and Jarque-Bera test are significant, GWR model could reduce model bias problem? If so, how can I check this in Arcgis?

Thank you and hope hear soon.

Number of Observations:      254          Number of Variables:                         5        
Degrees of Freedom:          249          Akaike's Information Criterion (AIC) [2]:    413.695641
Multiple R-Squared [2]:      0.385500     Adjusted R-Squared [2]:                      0.375629 
Joint F-Statistic [3]:       39.051951    Prob(>F), (4,249) degrees of freedom:        0.000000*
Joint Wald Statistic [4]:    132.866566   Prob(>chi-squared), (4) degrees of freedom:  0.000000*
Koenker (BP) Statistic [5]:  42.006323    Prob(>chi-squared), (4) degrees of freedom:  0.000000*
Jarque-Bera Statistic [6]:   20.054571    Prob(>chi-squared), (2) degrees of freedom:  0.000044*