st: spmat: banded matrix from friendship list (nlist) + another -sppack- question

[László Sándor <sandorl@gmail.com>]

Hi,

I have another question about -sppack-. I would use it with Stata 11.2
on mac and unix.

My understanding of the documentation is that -spmat import ...,
nlist- does not automatically give me a banded matrix. Is this
correct?
http://econweb.umd.edu/~prucha/Papers/WP_spmat_2011.pdf

However, I have very many people in my data from a social network, so
I need to end up with a banded matrix somehow. My two related
questions would be:
-- Could I generate a banded matrix outside of Stata that -spmat- can
-spmat use-? I did not find much guidance for that in documentation --
if it's not there, where could I read more about this? No carefully
crafted text file would do, you need something binary?
-- Or if I attempt a monstrous -spmat import- in order to be able to
-spmat tobanded- right after that, is there any way to find a helpful
sorting to do an -spmat permute-? My graph theory fails me here. I can
analyze the graph extensively, I pre-digest it with networkx from
networkx.lanl.gov.

Finally, I learnt from Rafal's kind answer a week ago that -spmat-
definitely allows asymmetric matrices. However, my problem does not
allow a canonical spatial-AR application, as relevant second-order
(neighbors' neighbors) links are not simply a square of the first-oder
contiguity matrix etc. (FYI: I tried to explain it with the chronology
of peer effects last week: previous cellmates of my current cellmates
have an effect on me, by his future cellmates cannot -- though I will
have an effect on them.) I would still try to use -spreg- for what I
can. Can I use it considering only first-order links? (Maybe this is
impossible in an SPAR framework and the most I can use is -spmat
lag-?)

Thank you very much,

Laszlo
*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/