st: cross-classified random effect model with xtmelogit

[Mrs Helda Khusun <helda.khusun@uqconnect.edu.au>]

Dear satalister,

I am trying to fir a cross-classified random effect model for estimation of age, period and cohort effect of a longitudinal data, in which age are nested and cross-classified in time of survey and birth cohort. The outcome is whether or not a person is overweight. Each person is followed longitudinally and measured at three time periods.

Initially I used the following syntax:

. xtmelogit overwt0 agetrc year0 bcohdum ||_all:R.year0||bcohdum:||pidlink: if s_analysis==1, laplace

where: overweight is the outcome (0/1), agetrc(age, continuous), year0(period of survey, 0,1 and 2), bcohdum(dummy of birth cohort, 0-7)

And I get sensible result from that.

But when I change the laplacian approximation into:

. xtmelogit overwt0 agetrc year0 bcohdum ||_all:R.year0||bcohdum:||pidlink: if s_analysis==1, intpoints(2)

I got the following error message after the second iteration: "Hessian has become unstable or asymmetric"

What does it mean and what shoudl I do next?

Any feedback would be very much appreciated.

Best regards,
Helda


*
*   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/