re: st: binary_mediation - a clarification

["Ariel Linden, DrPH" <ariel.linden@gmail.com>]

Paula,

There are some basic posting rules for this listserve. Perhaps the most
relevant rules that apply to your post is (a) note where you got the program
(command) from. More specifically, -binary_mediation- is a user-written
program (search binary_mediation, all), and it has recently been updated,
and (b) you are asked to provide the specifics of your model, that is, how
did you use the command and what was the output? Nobody will be able to give
you a concrete response without knowing exactly what happened. In the wise
words of Jerry Maguire (1996), "help me to help you."

Ariel



Date: Thu, 25 Oct 2012 13:45:05 +0100 (BST)
From: Paula Arce <paulaarce@rocketmail.com>
Subject: st: binary_mediation - a clarification

Hi There, 


I have used the binary_mediation command with 500 bootstrap cycles.  The
output shows me that the proportion of the total effect mediated is equal to
27% (the direct x--> y regression also shows that x does predict y). 


However, the 95%CIs for the indirect effect contain 0 (CIs: -.01 to .14),
which then suggests that mediation is not occurring. 


I am slightly confused by these results - should I only look at the
proportion value if I find evidence that supports the presence of an
indirect effect?

Thanks,
Paula



*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/