st: gf type evaluator

[Klaus Pforr <kpforr@googlemail.com>]

Dear Listers,

I'm working on ml evaluator for panel data and have a question about the 
gf2 type evaluator in moptimize. The manual and help files say that gf 
has to produce the group-level likelihood, the group-level scores and 
the hessian. I'm confused about the scores, as the manual says that the 
score matrix is supposed to be a matrix L x K, with L being the number 
of independent elements, i.e. the groups, and K being the number of 
coefficients. Does S really expect the derivative with respect to the 
parameters, i.e. the linear combinations for each equation, or the 
actually the derivatives with respect to the beta coefficients. If the 
first is true, I do not understand how I implement observation level 
scores with a likelihood that is only defined at the group level.

kind regards

Klaus
--

__________________________________

Klaus Pforr
MZES AB - A
Universität Mannheim
D - 68131 Mannheim
Tel:  +49-621-181 2801
fax:  +49-621-181 2803
URL:  http://www.mzes.uni-mannheim.de

Besucheranschrift: A5, Raum A312
__________________________________


*
*   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/