Re: st: Testing whether the average partial effects are statistically different across groups?

[Maarten Buis <maartenlbuis@gmail.com>]

On Wed, Aug 31, 2011 at 1:13 PM, natasha agarwal wrote:
> Thank you very much for your kind assistance.
>
> Can this also be performed after xtprobit, re?

As long as you are willing to fix the group level random variable at
the mean, but that does not fit well with the logic of average
marginal effects.

If you really want to report your results as only one marginal effect
per variable than you are much better of using a linear probability
model, as that model assumes that there is only one marginal effect
per variable. If that model does not fit well to your data, than
fitting an -(xt)probit- and reporting one marginal effect per variable
afterwards is not going to solve that problem. The "solution" that
-(xt)probit- and -(xt)logit- offers is that it makes the regression
lines non-linear, i.e. the marginal effects are allowed to change from
individual to individual. If you just report one (average or
otherwise) marginal effect, than you are undoing the very fix that is
implied with your -(xt)probit- or -(xt)logit- model. If a linear
probability model does not fit, than your best choice is to forget
about -probit- and move -logit- and interpret odds ratios.

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/