If you do not have a memory limit, then there is a good chance that it will
finish running -- at some point. The problem with making relogit run faster
is twofold:
1) The correction procedures are programmed in R, not C, so they are slower
than they should be. If you are handy with C, you could try translating the
R code (pretty clear, if I do say so myself) into C or C++ and then calling
those functions from Zelig.
2) The estimation procedure (both for the base logit and the relogit
correction) rely on serial rather than parallel processes. So having a grid
isn't going to help you (except in that you're not going to run out of
memory).
Yours,
Olivia
On Tue, Dec 21, 2010 at 9:44 AM, E. Lin <elin(a)iq.harvard.edu> wrote:
I'm trying to run a relogit model thru zelig with
1.4 million observations,
using weights, with about 15 covariates. Has been running for quite some
time. Is this just way to big to run, or is there some way to run this with
large data? Memory should not be a problem, as I'm running it on a grid
with about 50 gb of allocated RAM.
Any guidance would be much appreciated. thanks!
EXL
--
------------------------------
ERIC LIN
elin(a)iq.harvard.edu
------------------------------