No, the ssm do file breaks before gllamm is called. I now suggest you
set trace on and then re-run the ssm command to trace where in the ssm
program the error occurs, as in the mentioned FAQ. I suspect a small
problem with the ssm code for the pweight option but you need proof.
Also, can you reproduce the error for a more familar Stata dataset,
e.g. auto.dta? The idea is to reproduce the problem without confusing
extra information, to make it easy for the programmers to fix it if
need be.
Anders Alexandersson
andersalex@gmail.com
On 9/29/06, A Das <kerbo2004@yahoo.com> wrote:
Thanks, Anders. Yes, I meant sampling weights. This is
what happens when I do as you suggest-
. ssm emoprobb age harassd puberty11 firstvi ktouch
usborn2 [pweight=rweight], s(harassd = age edimp
milserved) adapt q(16) family(binom) link(logit)
commands
*-------------------------------begin
do-file-----------------------------------
* Select sample
mark touse [pweight= rweight]
markout touse emoprobb
gen long wgt = rweight
* deal with probability weights
gen _pwt2 = wgt
; invalid name
r(198);
So a problem with the underlying glamm code, then?
--- Anders Alexandersson <andersalex@gmail.com> wrote:
> On 9/28/06, A Das <kerbo2004@yahoo.com> wrote:
> > I'm trying to use the new ssm wrapper (Stata
> > Journal (2006), 6, 3, pp. 285-308) to fit a logit
> > model with a single endogenous dummy. The model
> does
> > not work when I include population weights
> (although
> > it should: p.292), but works when I do not. Here's
> the
> > model and the error message:
> >
> >
> > ssm emoprobb age harassd puberty11 firstvi ktouch
> > usborn2 if xf==1 [pweight=rweight], s(harassd =
> age
> > edimp milserved) adapt q(16) family(binom)
> > link(logit)
> >
> > ; invalid name
> > r(198);
> >
> > Again, the same model runs properly when I remove
> the
> > population weights. Any help would be greatly
> > appreciated.
>
> With "population weights" I assume you mean sampling
> weights, since
> you refer to pweight. I remember that gllamm instead
> assumes
> pweight(varname) but that's maybe irrelevant here.
> Is your gllammm
> installation up-to-date? To have more control of
> what's going, I would
> make a backup copy of the original dataset, then use
> the commands
> option, as described in section 6.4 of the article.
> If you still get
> an error message, at least you will now if the error
> message is for
> the underlying gllamm code or for the commands
> before gllamm is
> called.
>
> More generally, it is a good idea to debug a
> problematic command with
> -set trace on- beforing issuing the command; see the
> FAQ at
> http://www.stata.com/support/faqs/lang/debug.html.
>
> Anders Alexandersson
> andersalex@gmail.com
> *
> * For searches and help try:
> *
> http://www.stata.com/support/faqs/res/findit.html
> * http://www.stata.com/support/statalist/faq
> * http://www.ats.ucla.edu/stat/stata/
>
__________________________________________________
Do You Yahoo!?
Tired of spam? Yahoo! Mail has the best spam protection around
http://mail.yahoo.com
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/