I have a follow-up question concerning the -spatreg- command and
non-symmetric spatial weights matrices and would be very obliged for any
help on this.
I use -geneigen- command to generate an N by 4 matrix of eigenvalues
(containing the modulus, imaginary part, real part and cumulated modulus)
for my non-symmetric spatial weights matrix. I then indicate the name of
this N by 4 matrix in the brackets following the eigenval option in
_spatreg_. The command works and I get a table of results on my screen but I
am not sure whether the computation is done correctly since -spatreg- is
designed to be used with symmetric weights matrixes that produce an N by 1
matrix of eigenvalues. I am also not sure whether I should use all four
columns (i.e., modulus, imaginary, real or cumulated modulus) or only some
of them.
Thank you very much in advance.
Kate
-----Original Message-----
From: owner-statalist@hsphsun2.harvard.edu
[mailto:owner-statalist@hsphsun2.harvard.edu] On Behalf Of Kate Ivanova
Sent: Wednesday, March 29, 2006 4:49 PM
To: statalist@hsphsun2.harvard.edu
Subject: RE: st: Spatial lag models
Thank you very much, Nick and Kit, for your replies. I really appreciate
your help.
As for the theory on spatial models and panel data, Anselin in his book
"Spatial Econometrics: methods and models" (1988), for example, suggests the
spatial Seemingly Unrelated Regression (SUR) model that consists of an
equation for each time period which is estimated for a cross-section of
spatial units (coefficients vary for each time period but are constant
across space). But I guess there aren't any tools in Stata yet to estimate
such models.
Thanks again!
Kate
-----Original Message-----
From: owner-statalist@hsphsun2.harvard.edu
[mailto:owner-statalist@hsphsun2.harvard.edu] On Behalf Of Nick Cox
Sent: Wednesday, March 29, 2006 6:57 AM
To: statalist@hsphsun2.harvard.edu
Subject: RE: st: Spatial lag models
David Drukker points out that I overlooked
-natrix eigenvalues- as another way to do it.
Somehow or other that command remained
latent
in my memory and I was unable to
root
it out. That's, I find,
characteristic
of some commands, regardless of their inherent
value.
Nick
n.j.cox@durham.ac.uk
Moral: "To the vector belong the spoils" (Norton Juster)
n j cox
> The eigenvalues would ideally come from -spatwmat-,
> which was written for Stata 7, and uses -matrix symeigen-,
> hence the restriction you report.
>
> In Stata 7 or 8, there is a -geneigen- program by Kit
> Baum on SSC that doesn't require symmetry.
>
> . findit eigenvalues
>
> would have pointed you towards that.
>
> In Stata 9, Mata would, I presume, be the best way to do it.
>
> As for spatial data and panel models, does the _theory_ exist
> even?
> Kate Ivanova
>
> I need to estimate a model with a spatially lagged dependent
> variable and
> have downloaded a Stata add-on called spatreg. However, I
> cannot use it
> since my spatial weight matrix is not symmetric and the program cannot
> compute the eigenvalues.
>
> I would appreciate it very much if anyone could let me know
> if there are any
> ways in Stata to estimate a spatial lag model with a
> non-symmetric spatial
> weight matrix or any procedures to compute the eigenvalues for a
> non-symmetric matrix.
>
> I am also interested if there are any commands to estimate a
> spatial lag
> model using panel data.
>
> *
> * For searches and help try:
> * http://www.stata.com/support/faqs/res/findit.html
> * http://www.stata.com/support/statalist/faq
> * http://www.ats.ucla.edu/stat/stata/
>
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/