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: Re: st: MIXLOGIT: marginal effects |
Date | Thu, 9 Feb 2012 09:51:09 +0100 |
On Thu, Feb 9, 2012 at 3:08 AM, Nick Cox wrote: > I can readily believe in Kit's colleague's counterexample without even > seeing it. But if sometimes being quite the wrong model to fit is a > fatal indictment, then nothing goes. > > I was responding to Clive's statement "There is no justification for > the use of this model > _at all_ when regressing a binary dependent variable on a set of > regressors." I think that is too extreme. I can't readily imagine many > situations in which I would prefer a linear probability model to a > logit model, but I still think it's too extreme. One situation I can imagine where the linear probability model might be useful and unproblematic is what econometricians call a difference in difference design: You have two groups --- one experienced the treatment and one did not --- and two measurements --- one before the treatment and one after. The change in the outcome within the control group is thought to represent the change that would have occurred if the treatment was not administered and the effect of the treatment is the change that occurred on top of that "natural" change. So it is just a regression with two indicator variables and their interaction, and the interaction effect is interpreted as the effect of the treatment. There are many ways in which this can go wrong (e.g. where did the control group come from?) and the results can easily be over-interpreted, but I can still imagine many situations where a difference in difference model is a meaningful and useful strategy. Note that this is a fully saturated model, and as I showed earlier <http://www.stata.com/statalist/archive/2012-02/msg00351.html>, a logit or linear probability model give exactly the same predictions and both give valid inference. So the choice between the two is just a matter taste. No model works well in all situations, it is just a matter of having a large enough toolkit so you can choose the right model for the job at hand. Hope this helps, 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/