Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.
From | Maarten Buis <maartenlbuis@gmail.com> |
To | statalist@hsphsun2.harvard.edu |
Subject | Re: st: ZOIB procedure |
Date | Tue, 20 Sep 2011 12:20:22 +0200 |
--- On Tue, Sep 20, 2011 at 12:02 PM, Prerna S wrote: > Maarten, here is what I understand of the marginal effects. One can estimate > > a) mfx, predict (pr) - this is the marginal effects of the proportion > b) mfx, predict (pr0) - marginal effects of the dependent variable = 0 > c) mfx, predict (pr1) - marginal effects of the dependent variable = 1 > d) mfx, predict (prcond) - marginal effects of the dependent variable = (0,1) > > I am interested in b) and d) so I want to be clear on whether I have > this right. So are you suggesting that of these 4 options a) is the > best route to take whereas the remaining 3 are difficult to explain? No, you just need to define clearly what you are presenting and why. In my experience in most applications people just don't care about b-d, and in that case you obviously should not present them. If your problem is different than that conclusion does not apply. In your case you might actually want to look at the exponentiated coefficients as an alternative, because you seem to want to look at the conditional effects rather than the unconditional. They are not easy (look at -help betafit- for a discussion and example of relative proportion ratios), but they have the advantage of being a complete and concise representation of the model, while marginal effects are only approximations. Also a lot can be learned by trying to reconcile the results from marginal effects with these ratios. -- Maarten -------------------------- Maarten L. Buis Institut fuer Soziologie Universitaet Tuebingen Wilhelmstrasse 36 72074 Tuebingen Germany http://www.maartenbuis.nl -------------------------- * * 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/