Re: st: does margins command obviate concerns that inteff solves?

[Maarten Buis <maartenlbuis@gmail.com>]

On Fri, Dec 9, 2011 at 5:39 PM, Doug Hess <douglasrhess@gmail.com> wrote:
> Thanks, Maarten. My question, perhaps, is due to terminology. By
> adjusted predictions I mean I want to tell the story using percents.
> In my case, the percent of households that are food insecure (binary
> outcome). To me that is marginal effects.
>
> For instance:
>
> Model 1:  the effect on the outcome of household race, income, and education
> Model 2: the effect on the outcome household race, income, education,
> and educationX race
>
> If I want to find the significance of interaction using -margins,
> contrast-, do I need to also use inteff?

If you want a story in terms of risk differences than you should use a
model that is based on risk differences, i.e. the linear probability
model (-regress- with the -vce(robust)- option). If that model does
not fit to your data (e.g. predictions outside the range 0-1(00)),
than risk differences/marginal effects just aren't an appropriate
metric for your problem. You are not going to solve that by estimating
a -logit- or -probit- and than report marginal effects. If you report
one (average) marginal effect per parameter than you have basically
fitted a linear model on top of your non-linear model and reported the
coefficients of your linear super model. So now you are back at the
linear model that did not fit.

-- 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/