re: st: testing mediation in quantile regression

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

Hi Sam,

My question is why? What is your reasoning for using quantile regression? 

The R program you reference -mediation- is written by the same folks that
wrote the Stata program -medeff- (findit medeff). In neither program is
quantile regression an option. The R program does allow you to use general
additive models (GAM), as well as ordinal models. The Stata version does not
have those options.

I have not seen any papers (or books) discussing the use of quantile
regression in mediation analysis. So this would suggest to me that unless
you're trying to break new ground here, you may want to reconsider your
choice of models.

Ariel


Date: Thu, 17 May 2012 17:41:48 +1000
From: Samantha Batchelor <s.batchelor@griffith.edu.au>
Subject: st: testing mediation in quantile regression

Hi,

Can anyone help me with code to test mediation when running quantile
regression?  R has a package that does it but the similar one for
Stata (called mediation) doesn't handle quantile regression.
  For example, the mediator and outcome model below.

xi3: sqreg readingsc i.gender i.ethnicity i.mothered q9, q(.25 .5 .75)
reps(500)
xi3: sqreg comp readingsc i.gender i.ethnicity i.mothered*q9, q(.25 .5
.75) reps(500)

I've tried using R as well but I cant figure out how to deal with
interactions in the mediation code (or even how to get the effect of
q9 at each level of mothered after I have a significant interaction,
that is, the equivalent of using q9@i.mothered in Stata) so if anyone
knows quantreg in R and wants to point me in that direction that would
be ok too.

Thanks in advance,

Sam Batchelor


*
*   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/