Re: st: Interpreting 3 way dummy interaction with margins

[Maarten Buis <maartenlbuis@gmail.com>]

On Tue, Feb 7, 2012 at 1:57 PM, Colleen Nugent
<cnugent@sociology.rutgers.edu> wrote:
> I'd like to chime in to this conversation.  My apologies if this is a "dumb" question:  Maarten, if your tip for leaving out the reference categories is used with logistic regression and exponentiated coefficients, how is this different than the "predicted odds" generated with the margins command and "expression(exp(xb()))"?

It isn't different, as is shown in the example below:

*---------- begin example -------------
sysuse auto, clear
recode rep78 1/2=3
logit foreign i.rep78, or
margins, expression(exp(xb())) over(rep78)
logit foreign ibn.rep78, or nocons
*----------- end example --------------

> Actually, I know that it is different because I ran it, but I don't know why it should be different.  In my case the odds without the ref cat are all under 1.00 for the groupings on my interaction variables, but with the predicted odds using margins, they are all over 1.00.  So I guess I'm not understanding something here about what margins does.

With interactions you need to be careful and make sure that the
coefficients mean what you think they mean. There are typically
different ways of setting such models up that lead to very different
numbers for the coefficients even though the models are equivalent.
The devil is here really in the detail.

--------------------------
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/