Statalist


[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

st: 2 part model with Negative binomial


From   "Holland, Margaret" <Margaret_Holland@URMC.Rochester.edu>
To   <statalist@hsphsun2.harvard.edu>
Subject   st: 2 part model with Negative binomial
Date   Tue, 3 Nov 2009 11:56:23 -0500

I am attempting to test a mediation model, with an outcome that is a
count variable (Y), a mediator (M) that is continuous (but could be
dichotomized), and an independent variable (X) that is either
dichotomous or continuous. I would like to test the indirect path, which
can be considered a 2-part model:

Y=M+covariates
M=X+covariates

I am considering negative binomial as the best model for my outcome (Y),
although there are many zeros (70%). Zero-inflated negative binomial or
ZIP generally do not converge well for these data. For the second
equation, OLS should be sufficient or perhaps logit or probit.

I have not been able to find something already written in Stata to
estimate this. I am familiar with user-written MLE, but fear this would
get quite complicated if I want to included correlated errors, which
seem likely.

Any thoughts?
Thank you,
Maggie

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/



© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index