Dear all,
I am programming a ML code for a FIML endogenous switch count model as in
Terza (1998). The code uses hermite quadrature for calculating
the log-likelihood. Unobserved heterogeneity for the count process, e, is
allowed to be correlated with unobserved heterogeneity for the switch, u,
and var(e) is estimated along with the correlation coefficient of e and
u, rho. Rho is constrained to be between -1 and 1. I am using a d0
method.
Everything seems to be ok, i.e. ml check does not complain and no
problems using search. But after some iterations rho hits 1, and though
the code sets rho to .99 each time rho>=1, rho goes out the admissible
range. Does anyone have an idea what is going on here?.
I would like to test my ML code using simulated data in order to discard
any problem with the data. Could you help me to create a simulated count data
set with one dummy endogenous variable?
Many thaks,
Alfonso Miranda
Warwick University.
*
* 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/
