On Wed, Nov 25, 2009 at 10:44 AM, Marco Alfano <[email protected]> wrote:
> Dear Statalisters,
>
> I am trying to estimate a multilevel structural equation model using gllamm. Basically I am interested in the effect of a latent on an observable variable. This latent variable in turn is formed out of three other latent variables and I have a number of fallible measures for each of these three. I have coded this model and have pasted the relevant parts at the bottom. It runs ok for some subsamples; for others, however, the model does not converge (with the messages “discontinuous region encountered” or “backed up” or “not concave”). Does anyone have any advice? Any help would be greatly appreciated.
I don't think you have the right -gllamm- syntax. For every indicator
of a latent variable, you should have a line in your data, so in the
end it should have three levels, with mothers, kids, and indicators
nested within kids. Also, with two integration points, you won't get
any results reliable numerically.
--
Stas Kolenikov, also found at http://stas.kolenikov.name
Small print: I use this email account for mailing lists only.
*
* For searches and help try:
* http://www.stata.com/help.cgi?search
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
