Help on negbink          J. Hilbe 6-15-94
---------------

Log-link negative binomial regression

Command line:

   ^negbink depvar predictors [if] [in] [<fweight/aweight>=var], k(#)
   ^[lt] [it] [<eform>] o[<offset>] [<nol>] [<nocon>] [<expec(#)>] [<resid>]
   
k(#)      = value for alpha(k); default=1 (geometric regression)
lt(#)     = tolerance for convergence
it(#)     = number of iterations
eform     = results in exponential form (IRR ratio)
o         = offset variable
nol       = no display of iteration log
nocon     = noconstant model
expec(#)  = number of iterations using Fisher scoring (default=1)
resid     = creates the following variables:
            _mu (fit)                     _lp (linear predictor) 
            _Pear (Pearson residual)      _Dev (deviance residual)
            _SPear (standardized Pearson) _SDev (standardized deviance)
            _Hat (hat matrix diagonal)

The value of alpha or k can be understood as the amount of overdispersion 
present in otherwise Poisson-type count data. Set k such that the value 
of the CHI2 dispersion following estimation approximates 1.0

For additional assistance contact by email: hilbe@@asu.edu