Date:         Wed, 2 Oct 1996 20:23:01 GMT
Reply-To:     Tracy Clark <Tracy.Clark@asu.edu>
Sender:       "SAS(r) Discussion" <SAS-L@UGA.CC.UGA.EDU>
From:         aqtlc@IMAP1.ASU.EDU
Organization: Arizona State University
Subject:      Proc Model Fairmodel

In order to learn the ins and outs of proc model I am trying
to convert the fairmodel detailed at
http://fairmodel.econ.yale.edu/ to sas and proc model.
The conversion has raised some issues that the little
examples in the manual can't answer and I was hoping that
there were some kind souls out there who had experience with
using proc model with a bigger model.
 
The question I have today is in regards to different instruments
for each equation under 2sls.
 
It seems to work to just have a separate fit statement for each
equation. That is the parameters are floating around somewhere
so that a solve command that references all the fitted vars has
something to chew on. Computationally is this sound?
 
With each equation being fitted separately  if an out= is added to any
fit command including the last one,the only var for which the actual
and fitted values differ is for the var named in the fit command.
Values for all the vars are dumped but the actual and fitted values
are the same. Is it possible to dump the fitted values some other way
or at least limit what is written to the dataset to the var in question
so that the fitted values can be merged later?
 
Thanks to anyone who can help.
 
I would also be interested in peoples experiences in using proc model
for larger models. The last time I tried to do a large scale model was
more than ten years ago and we did the data is SAS but the model in
TSP.