As many may know, Stata's -nbreg- command provides users with the ability to
model data using either:
NB-2 , nbreg y varlist, dispersion(mean) DEFAULT
NB-1 , nbreg y varlist, dispersion(constant)
NB-1 is sometimes called linear negative binomial because its variance
function is mu+alpha*mu, whereas the traditional NB-2 parameterization has a
quadratic variance function of mu + alpha*mu^2.
The command -gnbreg- allows parameterization of the negative binomial
heterogeneity or ancillary parameter, but only for the NB-2 model. One cannot use
gnbreg to parameterize the NB-1 heterogeneity parameter.
A few days ago I submitted a command to SSC to parameterize the
heterogeneity parameter for the canonical negative binomial, or NB-C. The command name is
-hcnbreg-. I have now submitted a command, called -hnbreg1- to parameterize
NB-1's heterogeneity parameter.
To install from the Stata command line, type
ssc install hnbreg1
or
ssc describe hnbreg1
then install where shown.
Note that, unlike -nbreg-, Stata's -gnbreg- command does not provide a
likelihood ratio test of the estimated model versus Poisson, which is a negative
binomial with a heterogeneity parameter value of 0. I have added this feature
to the hnbreg1 command, together with an AIC statistic. Both statistics
assist the user in assessing the comparative fit of the model. The help file
explains the other features and options.
Joseph Hilbe
************************************** See what's new at http://www.aol.com
*
* 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/
