Spatial Panel Data and Large Dataset

[Nikhos Ablos]

Hi !

I have a large panel dataset with 33 000 observations every year for 25 years on STATA 14.
I would like to use command such as xsmle to allow for municipality-specific trends, spatial autocorrelation and other things.

For that, I first need to compute a spatial weight matrix with a command such as spmap. spmap cannot compute a 33 000x 33 000 matrix, I therefore need to compute a banded matrix but xsmle doesn't allow to use banded matrix.

My questions are therefore:
- Is there another command than xsmle that allows to use banded matrix ?
- Is there another way to allow for municipality-specific trends, spatial autocorrelation, etc. than by computing a spatial weight matrix ?

Thank you in advance !

[Phil Bromiley]

You didn't get a quick answer. You'll increase your chances of a useful answer by following the FAQ on asking questions - provide Stata code in code delimiters, readable Stata output, and sample data using dataex. There are many relevant entries if you google this:

https://www.stata.com/statalist/arch.../msg00089.html
http://econweb.umd.edu/~prucha/Paper...%282013%29.pdf
http://wlm.userweb.mwn.de/Stata/wstaspwm.htm

[Nikhos Ablos]

Hi !
Thank you for your help. I have checked these references but it does not solve my problem I think.

I tried these following codes:

1/

spatwmat, stand name(mat1) xcoord(latitude) ycoord(longitude) band(0 10)

that return me matsize too small

2/

spmat idistance mat1 longitude latitude , id(id) norm(row)

SPMAT_idistance(): 3900 unable to allocate real <tmp>[35646,35646]
<istmt>: - function returned error
r(3900);

3/ I therefore decided to use banded matrix but I therefore cannot use xsmle anymore.

spmat idistance mat1 longitude latitude , id(id2) norm(row) banded btruncate(1 3)
spmat save name using mat1.dta
xsmle  depvar [indepvars] , wmatrix(mat1) model(sar)

xsmle does not support banded matrices
r(498);


Do you have any ideas ? Do you know other commands?
Thanks !

[Lom Newton]

Hello Nikhos Ablos, please were you able to find a solution to the "xsmle does not support banded matrices" problem? I face the same problem. If you have found a solution or an alternative, could you kindly share it?

Thanks.

Lom.