Statalist


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

RE: st: What multiple regression model for extreme distributions


From   "Verkuilen, Jay" <JVerkuilen@gc.cuny.edu>
To   "'statalist@hsphsun2.harvard.edu'" <statalist@hsphsun2.harvard.edu>
Subject   RE: st: What multiple regression model for extreme distributions
Date   Wed, 3 Feb 2010 13:19:44 -0500

Ronan Conroy wrote:

>>Then this sounds like a two-stage model. The first model predicts  
whether the family saves or not, the second predicts how much they  
save, conditional on their saving at all. I would be wary of  
predicting everything in a single model - the determinants of how much  
may not be the same as the determinants of whether.<<

My thoughts exactly!

And to echo Maarten's point, -zinb- may be a good option. It lets you do the kind of two stage estimation you probably need. Yes it's for counts but if you have continuous numbers, you probably won't hurt anything to round to integers, assuming that -zinb- checks for that. If that's offensive to you, it should be possible to create a ZI model in -gllamm-. 

*
*   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