strange results from OLS regression. 1.00000 probability

[asifai...@gmail.com]

Hello,

 

I am using GeoDa for spatial econometrics. I have excel file (with more than 8938 observations (rows) and 215 variables (columns). I saved this file in dbf format by using SPSS.

when I open this file in GeoDa it shows some missed values in the table and then I am unable to do regression as not all variables are present.

 

Later I changed the format from excel workbook file to Ms Excel (.xls). 

Now it shows all variables but then I am getting very strange results for simple OLS regression for example probability is 1.00000 for all variables and most of the results having  weird numbers.

Do you know why GeoDa produces results like this?

 

 

Variable            coefficient        std. Error                                     t-Statistic                                     Probability

LNA                    3.210924          5.590378e+012                          5.743661e-013                            1.00000

LNR                    -1.209391         5.025374e+012                          -2.40657e-013                            1.00000

LNA                    -0.5997587        2.657702e+012                         -2.256681e-013                           1.00000

 

 

Moran’s I is about 0.08

Another important point is I am unable to create Contiguity-Based Spatial Weights as the option is simply faded.

I can only do distance based with Euclidean distances by selecting x and y coordinates or

Otherwise I can select k neighbors I am using 5 neighbors but the results are same in both options.

 

 

Do you have any idea what to do in this situation? Or what is the actual problem here.

Thanks in advance

 

 

Best Regards

 

Asifa Iqbal

[Julia Koschinsky]

It sounds like you are using point data (XY coordinates). Earlier versions of GeoDa only did distance-based weights for points. If you download the latest version of GeoDa, it also does contiguity weights for points based on Thiessen polygons that the points are converted to on the backend.

Your regression results point to a problem in model specification, e.g. perfect collinearity. Take a look at the distribution of the variables in your model to narrow down where the problem lies.

Julia

--
You received this message because you are subscribed to the Google Groups "Openspace List" group.
To unsubscribe from this group and stop receiving emails from it, send an email to openspace-lis...@googlegroups.com.
To post to this group, send email to openspa...@googlegroups.com.
Visit this group at https://groups.google.com/group/openspace-list.
For more options, visit https://groups.google.com/d/optout.

--

************************
Julia Koschinsky, Ph.D.
Research Director
Associate Research Professor
Arizona State University
School of Geographical Sciences and Urban Planning
GeoDa Center for Geospatial Analysis and Computation
julia.ko...@asu.edu

http://geodacenter.asu.edu
http://www.facebook.com/geodacenter
http://twitter.com/GeoDaCenter

[asifai...@gmail.com]

Thank you very much. 

Asifa

[asifai...@gmail.com]

just to show that this is not the problem with my file. I downloaded the data avaialable at GeoDa center website and it is same faded option for contiguity weights. 

Kind Regards

Asifa

[asifai...@gmail.com]

Hello,

Now i am using new version of GeoDa (1.6.7) but the option for contiguity weights is still faded (I am unable to select queen or rook weights).

I have also tested another file with polygon data but the option to select queen or rook weights is unavailable. Do you have any idea what is the problem here and what to do?

Distance based weights creats GWT file. and this only allows me to run classic regression. I am unable to run spatial lag or error model. "Error: The Inverse matrix is ill-conditioned".

While using a polygon file, running spatial lag gives this message "only symmetric weights are supported for this operation please choose a symmetric weights file".

Thanks in advance

Asifa

 It sounds like you are using point data (XY coordinates). Earlier versions of GeoDa only did distance-based weights for points. If you download the latest version of GeoDa, it also does contiguity weights for points based on Thiessen polygons that the points are converted to on the backend.