You might want to take a look at the R package "mediation" which my
collaborators and I created. It uses a simulation approach to compute
the uncertainty. Here are some papers:
http://imai.princeton.edu/projects/mechanisms.html
Good luck,
Kosuke
---------------------------------------------------------
Kosuke Imai Office: Corwin Hall 041
Assistant Professor Phone: 609-258-6601
Department of Politics Fax: 609-258-1110
Princeton University Email: kimai(a)Princeton.Edu
Princeton, NJ 08544-1012
http://imai.princeton.edu
---------------------------------------------------------
On Aug 23, 2009, at 11:29 PM, Donald Braman wrote:
Dear Zelig Folk,
I have a problem that you may have dealt with: mediating effects.
Here's a very simple version of the problem:
We have mock jury data in which:
Verdict (V) ~ Factual Perception (FP) + Egalitarianism (E)
Simple, right? Except that E influences V both directly and
indirectly through an influence on FP. So we really have two
equations:
V ~ FP + E and FP ~ E
Since both V & FP are dependent on E, I should be able to simulate a
distribution of FP with E set to X *and* a distribution of V with E
set to X and FP set to sim(FP[E=X]).
The way I'm trying to do this by simulating 1000 values for FP with
E set to X. Now I have FPsim[1...1000]. Now I want to simulate V
by setting E to X and FP set to FPsim[1...1000].
So my questions are: Is that reasonable way to estimate direct &
indirect effects in theory? How would you go about doing it in
practice?
Don
-
Zelig Mailing List, served by Harvard-MIT Data Center
Send messages: zelig(a)lists.gking.harvard.edu
[un]subscribe Options:
http://lists.gking.harvard.edu/?info=zelig
Zelig program information:
http://gking.harvard.edu/zelig/