Statalist


[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: st: AW: computing interaction effects after xtprobit


From   Austin Nichols <austinnichols@gmail.com>
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: AW: computing interaction effects after xtprobit
Date   Fri, 27 Mar 2009 13:32:49 -0400

I.Roland and Martin Weiss <martin.weiss1@gmx.de>:

I don't see an easy solution--for each obs, the full interaction
effect is the cross-partial derivative of the expected value of y
(http://www.stata-journal.com/sjpdf.html?articlenum=st0063, p.155)
which in -xtprobit- depends on the (unobserved) random effect
(http://www.stata.com/help.cgi?xtprobit+postestimation
shows you can get the probability of a positive outcome, assuming that the
random effect for that observation's panel is zero, but not an
estimated random effect).

You could integrate over simulated random effects, I suppose...
averaging not only over obs in the sample but across many possible draws
of REs for each obs (using estimated sigma_u and rho to simulate REs;
see http://www.stata-journal.com/sj6-2.html for some helpful technique).

SEs on these "full interaction effects" would be a pain then--perhaps
you can treat the REs as missing data, and do multiple imputation from
the distribution given by estimated sigma_u and rho, then use Rubin's
formulas to calculate SEs.

No doubt there is a better strategy, but it is not immediately obvious to me...


On Fri, Mar 27, 2009 at 1:07 PM, Martin Weiss <martin.weiss1@gmx.de> wrote:
>
> http://www.stata-journal.com/sjpdf.html?articlenum=st0063, page 156 says
> that you can replicate anything -ssc d inteff- does via -predictnl-.
>
> HTH
> Martin
>
>
> -----Ursprüngliche Nachricht-----
> Von: owner-statalist@hsphsun2.harvard.edu
> [mailto:owner-statalist@hsphsun2.harvard.edu] Im Auftrag von
> I.Roland@lse.ac.uk
> Gesendet: Freitag, 27. März 2009 18:02
> An: statalist@hsphsun2.harvard.edu
> Betreff: st: computing interaction effects after xtprobit
>
> Hello Statalisters,
>
> I would like to estimate a random-effects probit that includes an
> interaction term between two continuous independent variables. After the
> command 'probit', 'inteff' can be used to compute the correct interaction
> effects. However, it does not seem to work after 'xtprobit'. Does anyone
> know how to compute interaction effects after 'xtprobit'?
>
> Thanks in advance!

*
*   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/



© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index