st: Is it possible to use Nelder Mead optimization with ml

["Woodruff, Robert T" <WOODRUFFR@battelle.org>]

Hello all,

I wrote an --ml-- d0 program to optimize a conditional likelihood that is a nonlinear function of the parameters.  I'm using a stereotype regression model for a matched case-control study (and hence couldn't use the existing implementations for unconditional stereotype models).  For the most part it works fine but for some datasets it has a tough time with convergence.  I've played around with the different optimization options (except BHHH) and initial values but it can't seem to find the top of the hill.  I know Nelder-Mead has been implemented in Mata's --moptimize-- function but I was wondering if anyone knew if I can use it with --ml-- too.  

Also, I can compute the gradient and Hessian explicitly but currently am being lazy by just letting --ml-- compute them numerically.  Do you think this would make much of a difference for the methods that use the derivatives?

Thanks,
Rob Woodruff
woodruffr@battelle.org

*
*   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/