Re: st: question on gllamm for discrete latent variable with one factor structure

[Stas Kolenikov <skolenik@gmail.com>]

On Mon, May 10, 2010 at 2:36 PM, Hey Sky <heyskywalker@yahoo.com> wrote:
> here is the message after the iteration stopped:
>
> numerical derivatives are approximate flat or discontinuous region encountered
> could not calculate numerical derivatives
> missing values encountered
> (error occurred in ML computation)
> (use trace option and check correctness of initial model)

Most likely, your model is not identified. From your syntax, it
appears that you are specifying the constant term twice: in the fixed
effects equation and in the random effects equation(s). Try -nocons-
option if you need the constant in the random effects, although make
sure it does make sense for your model. Your equations (-eq- commands)
must have names; Go(ul)d only knows how Stata interprets the empty
-eq-s in -gllamm- context.

> I am still not familiar with the gllamm options so that I think there must has something wrong
> in my code. I also have tried to do as what the example on gllamm manual chpater 5.2.1 p59
> did, it does not work either and give the same message as above.
>
> the model should be ok since fortran code gives satisfied result. and I cannot use stata since
> the data center does not have it and not allowed to install it either.
>
> here is what I tried and edu is a 0/1 dummy:
>
> gen cons=1
> eq: cons
> eq: edu
>
> gllamm dep indep, i(id) base(5) link(mlogit) family(binom) nip(2) ip(f) nrf(2) ///
>  eq(cons mu2) dots trace


-- 
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/