GLM command for zero truncated negative binomial regression

[Erin Mead-Morse]

Is it possible to use the GLM command for zero truncated negative binomial regression? I believe the code for negative binomial regression is glm x y, family(nbinomial) link(log), but I do not know how to indicate truncation. Thank you for your assistance.

[Jeff Wooldridge]

I don't think that's possible, as the GLM command restricts attention to the linear exponential family of distributions. If you're mainly interested in the effects on the conditional mean I would strongly recommend using family(poisson), as it's a fully robust estimator. Are you interested in "truncation" because you have a lot of zeros?

[John Mullahy]

Jeff's point in #2 gets to the crux of the issue. But if you are set on estimating a truncated negative binomial model have you looked at the tnbreg command?