Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.
From | paulasim@portugalmail.pt |
To | statalist <statalist@hsphsun2.harvard.edu> |
Subject | st: Problems in ztnb and nbstrat estimation |
Date | Wed, 13 Apr 2011 16:41:50 +0100 |
Dear stata users, I am using STATA 11 version updated yesterday. I am dealing with count data concerning recreational visits behaviour. It was obtained by the administration of an on-site survey. I have 212 observations. When I use nbreg there are no problems, coefficients present expected signs and the over dispersion parameter (alpha) is about 1.26. Problems arise when I try to implement the zero truncated negative binomial model or the negative binomial with endogenous stratification. In both: - the over dispersion parameter (alpha) becomes huge, the ln(alpha) is about 20; - the constant term increases, in absolute value, but is not significant; - predictions are completely wrong with an average value of 0.0000... The nbstrat output reports that alpha is not significant. Is the number of observations excessively low? Had this happened before to anyone? Thank you very much, Paula Simões * * 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/