What you can do is to set up a random coefficients model where the
latent variables are loaded onto the constant and your time variables,
similar to section 3.2 of the 2001 version of -gllamm- manual:
gen one = 1
eq const : one
eq time : year
gllamm whatever, i(id) nrf(2) eqs(one year)
If you have say 10 or so time periods, I would model the time trend by
splines, although that would increase complexity of the factor part of
the model quite substantially.
Stas
On Sat, 30 Oct 2004 21:41:12 -0700, Shige Song <shigesong@gmail.com> wrote:
> Dear Colleagues,
>
> I am trying to estimate a multilevel growth model using GLLAMM. Since
> I have closely spaced observations for each individuals (2 years), I
> am a little concerned about the possibility that the error terms are
> somewhat correlated. I know there are two ways to handle this, one can
> either impose a structure on the error terms (AR1, for example), or
> one can model the error terms as function of age or time. Now I know
> that GLLAMM cannot estimate the first class of models, my question is:
> can GLLAMM estimate the second class of models? Are there any worked
> examples on this? Thank you very much!
>
> Best,
> Shige Song
> Institute of Sociology,
> Chinese Academy of Social Sciences
--
Stas Kolenikov
http://stas.kolenikov.name
*
* 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/