st: Mixed ordered confusion

[Lucas <lucaselastic@gmail.com>]

So, a little while ago Scott Baldwin and I discussed on statalist the
meaning of noconstant at level 2.  The example he provided was of boys
and girls nested in schools.  So, here's the problem.  Assume you have
boys and girls nested in schools, and the dependent variable is
ordered (say, # of Advanced courses taken in Math and Language--0, 1,
2).

If the variable were interval-level test scores, and score a year
earlier is "oldscore", one might model this as:

mixed test_score boy girl oldscore, noconstant || school: boy girl, nocons ;

If one thought school-level Z1 made a (possibly different) difference
for boys and girls one would write:

mixed test_score boy girl oldscore boy*Z1 girl*Z1, noconstant ||
school: boy girl, nocons ;

This is all well and good.  BUT, how to specify a similar model for
ordered outcomes?  One might think the code would be:

meologit AdvancedCourses boy girl oldscore boy*Z1 girl*Z1, noconstant
|| school: boy girl, nocons ;

but one cannot use the noconstant option for level-1 for ordered
variables.  I tried fixing the cut1 score at a chosen value, and
forcing retention of collinear terms, but this did not work.  Any
ideas?

Thanks a bunch.
Sam

PS--It is probably computationally difficult for programmers to invoke
the option, but it would be if users could specify whether they want
to estimate thresholds and force an omitted category to be zero or,
instead, estimate category adjusted means and force one threshold to
be zero.

Anyway, again, thanks a bunch for any assistance anyone can offer.

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