I have a question about what the inverse link function is for a ZIP and ZINB model to be able to generate manual predictions from given a set of the linear predictors for a particular model.
In STATA it's easy to predict, but our ultimate goal is to manually predict in other software applications given a suite of covariates. I've used this before with LOGIT models easily using the inverse link, but have been unable to determine if there is a simple inverse link for ZIP and ZINB because of the conditionality of the zero part of the models.
I did find a previous posting here that had code for manually predicting following a ZINB model [ posted by Fieveson http://www.stata.com/statalist/archive/2007-03/msg00531.html] and we have used this and it worked, but we noted a few problems that made me wonder if there isn't a one step inverse link to use in predictions. Our problem was that predictions from this ZINB model were quite poor relative to other measures of model fit, which were reasonable. In speaking with others, it seems that predictions are often not great with these ZIP models, but I wanted to clarify if there was an inverse link function equation for ZIP and ZINB like the other GLM models.
Thanks for any help.
Sincerely,
Mark Hebblewhite
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Assistant Professor
Wildlife Biology Program
University of Montana
Missoula, MT, 59812
Email: [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/
