I am interested in your problem. I suspect that the log-likelihood formula
used in nbinlf.ado for estimating the zero truncated negative binomial is OK.
It's not a difficult formula and it was checked many times prior to
distribution. I am wondering if the problem you are experiencing relates to the older
code in which the program was written. I believe it was version 5 or 6. I
had to entirely re-write my censored Poisson regression algorithm using version
8 code in order to add certain survey capabilities to it, including
pweights. I renamed it cepois, to distinguish it from the 1997 version, cenpois, and
later posted it to the Boston SSC site. I'll be happy to help with your
project of modifying the code to allow endogenous stratification. I suspect,
though, that it will entail a re-write to version 8 code. Or we can see what
capabilities version 9 has to make the project easier.
If you are interesting, contact me privately and we'll go from there.
Joe Hilbe
====================================
I attempted to modify Joseph Hilbe's 0-truncated negative binomial routine
(nbinlf.ado likelihood function) to accomodate for endogenous stratification
(e.g. on-site sampling), however I am running into problems during the
maximization process; probably inherent in the trnbin0.ado file which
performs
the estimation. Has anyone else attempted to do this? Would anyone be
able to
help?
Cheers,
Craig Meisner
-------------------------------------------------------------
Craig Meisner
The World Bank
Development Economics Research Group
Infrastructure and Environment
MSN MC2-205, 1818 H Street, NW
Washington DC 20433
Tel: 202-473-6852, Fax: 202-522-3230
E-mail: [email protected]
*
* 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/
