[Openspace] Regression with selected Obs.

[Jochen Mistelbacher & Sarika Sharma]

Dear all,

as a "Newbie" to GeoDa, which I unfortunately discovered only recently, 
I would like to pose a question regarding running a spatial-lag-MLE only 
on selected areas of a polygon-shapefile.

The Setting: I have a shapefile made up of more than 500 areal units 
(districts in this case), a dependent variable and multiple independent 
variables. To account for spatial autocorrelation, I run the 
spatial-lag-model on the whole set.

The Aim: I want to compare the coefficients for different regions of the 
whole set, and not like the "Geographically Weighted Regression" for 
each observation.

The Problem: I would like to run the spatial-lag-model using only 
selected areal units (in this case districts belonging to certain 
states). As of course these selected units have neighbours outside the 
selected area, simply creating new shapefiles and weights wouldn`t solve 
the problem, as then the peripheral areas would not have neighbours and 
the weights would not have a "true" value". If  I would include the 
first-order-neighbours in a new shapefile, so that their "influence" is 
taken into consideration , then of course the selected area would not be 
strictly limited to the next higher administrative boundary (here states).


Question: Is there a way to deal with this problem?


And as I am already on it, a most probably rather stupid question on 
MLE: Can I calculate the "Beta-coefficients" using MLE?

Thanking in advance for good advice!!

Jochen Mistelbacher

Dept. of Geography
University of Mannheim
Germany
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://agec221.agecon.uiuc.edu/pipermail/openspace/attachments/20040811/1afd27d7/attachment.htm