Monday, March 06, 2006

Re: Re: st: marginal effects for ZINB with interactions and squared terms

>interactions and squared terms > >Have you read >http://www.stata.com/support/faqs/stat/mfx_dydx.html >http://www.stata.com/support/faqs/stat/mfx_interact.html >yet? > >A good place to start.

Austin, thanks a lot for the links. However, the problem I've found is that mfx and margeff do not recognize interactions and squared terms, while nlcom and predictnl do account for that but need to be fed with the derivative of the function used in the model. If it is a logistic (for logit) or normal (probit), it is easy, but a ZINB is a nested model, and I don't know how to analytically compute it (and I could not find any source about it). Do you know how I can "skip" this problem? (if I'm not wrong, stata uses a numerical procedure, so that it could calculate the derivative of virtually whatever function). Thanks, Francesco

> >On 3/3/06, Rullani Francesco <rullani@sssup.it> wrote: >> I am estimating a model with interactions: >> a+b(x_1)+c(x_2)+d(x_1*x_2) >> and squared terms: >> e+f(z_1)+g(z_1)^2 >> using a zero-inflated negative binomial regression >>(ZINB) >> because of overdispersion and excess of zeros in the >>data. >> Does anyone know how I can obtain marginal effects and >> their standard errors, given that mfx and margeff do not >> "know" that I have interactions and squared terms? >

* * 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/


Tag:


Links to this post:

Create a Link



<< Home

This page is powered by Blogger. Isn't yours?