On Thu, 9 Jun 2005, fdmplimo(a)usp.br wrote:
> Dear professor King
>
> My name is Fernando Limongi. I work with Argelina Figueiredo taht
> recently talked with you about applying your solution to the ecological
> problem to Brazilian electoral data. I am using eziwin to get the
> estimates of split or join vote on Brazilain elections. The program runs
> only up to 3,000 or 4,000 cases. When I try with more cases, the program
> does not run. In the software documentation there is no reference to
> limitations due to the number of cases. Is there such a limitation? If
> yes, Is there something that can be done to circunvent this limitation?
>
> I am trying to run the data on EI but Gauss is not very popular in
> Brazil, so no one has a copy of the program. I tried to adapt your
> routine with some others softwares but always fial for the missinf of
> some of the programes invoked. With regard EI, Gauss is the only
> software that runs it?
>
> Thank you very much for your time
>
>
> Fernando Limongi
> Political Science Department at USP
>
>
You probably have just a hardware limitation; if you used a machine with
more RAM, odds are it will work. But there are workarounds. The key is
that the basic model only estimates 5 parameters in order to get you
precinct-level estimates for all n precincts. The extended model won't
usually have many more parameters.
That means you are using a lot of data to estimate those parameters and
you could use less without much cost. There are several options. The
easiest is to set _EselRnd to select (randomly) some fraction of the data
in the estimation stage; you'll still get estimates for all quantities of
interest at the precinct level. The other option would be to run the
whole analysis separate for different subsets (such as regions) of your
data. That too would work and would add some flexibility, allowing the
model to fit the data better.
Gary
---
Gary King
Institute for Quantitative Social Science
Harvard University, 34 Kirkland St, Cambridge, MA 02138
http://GKing.Harvard.Edu, King(a)Harvard.Edu
Direct 617-495-2027, Assistant 495-9271, eFax 812-8581
-
EI mailing list served by Harvard-MIT Data Center
Subscribe/Unsubscribe: http://lists.hmdc.harvard.edu/?info=ei