[Openspace] more questions...
syabri at uiuc.edu
Tue Dec 16 13:02:29 CST 2003
(1) If you look at the GAL weight file, you're right, the GAL file
shows only the list of neighbors for each observations.
Geoda then reads and stores that file into a 'sparse' matrix.
This sparse matrix will get standardized during the computation,
for examples, the computation of Moran's I, Lisa Map, etc.
However, from the user perspective, it seems that
the weights is already raw standardized.
(2) The 'treshold' will be fixed in the next version. thanks.
(3) Geoda can't calculate or construct a distance weight matrix for
a lower bound > 0, yet. So, the threshold distance in Geoda is
specified as follow: the lower bound is always set
to zero, the upper bound is set by the user (at least >= its cutoff point).
(4) The distance information created by Geoda is not the inverse distance.
The distance information in the distance weight file (GWT file) created
is the distance between pair of objects within the given threshold.
For now, although geoda can create and read a distance weight file,
it treats the distance weight file as a contiguity matrix.
I hope these will help.
----- Original Message -----
From: Coro Chasco Yrigoyen
To: Monica Palaseanu-Lovejoy
Cc: Openspace at agec221.agecon.uiuc.edu
Sent: Tuesday, December 16, 2003 11:18 AM
Subject: Re: [Openspace] more questions...
The weights are already raw standardized.
I'm not completely sure about this point because when calculating contiguity matrices, they're not row-standardized (if you have a look at the .GAL file with e.g. Notepad).
The word "threshold" is
Yes I know. But in GeoDa it is not correctly writen; it says "treshold".
, and you yourself answered the third question. How
did you calculated the expected spatial lag for 50 km and 175 km
to compare with what GeoDa does? if you use different thresholds
you will certainly get different spatial lags.
Yes I know. But they are completely different and sounds strange to me. I can re-compute it with SpaceStat to be completely sure and I'll tell you.
I don't know about computing an inverse distance matrix in GeoDa
but if you look into R - a free statistical software
(http://cran.r-project.org/), there is a package called spdep which
reads data from GeoDa like the weight matrix.
Thanks a lot. But I can calculate these matrices with SpaceStat. The question is why GeoDa cannot do that.
Coro Chasco Yrigoyen
Profesora Dpto. Economía Aplicada
Directora del Área de Economía Espacial Microterritorial
Instituto Lawrence R. Klein
Universidad Autónoma de Madrid
Telf.: ++34 91.497.42.66
E-mail: coro.chasco at uam.es
Openspace mailing list
Openspace at sal.agecon.uiuc.edu
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the Openspace