Re: st: Direct and indirect of intrument variables

[Stas Kolenikov <skolenik@gmail.com>]

As a simultaneous equations model, this system is not identified. I
don't think people in mediation literature ever bother about this, or
even understand this, but this means that even if you come up with
some sort of estimates from your two or three stage procedure, they
won't be trustworthy.

On Tue, Mar 29, 2011 at 12:05 PM, Garriga Rubio  Helena
<hgarriga@ethz.ch> wrote:
> Hello,
>
> I'm trying to do a two-step regression manually and then analyze the direct and indirect (mediator) effect.
>
> (1) IV --> MV1
> (2) IV --> MV2
> (3) MV1, MV2 --> DV
>
> After the regressions, I compute the standard errors accounting for the inclusion of a predicted regressor through the steps suggested in Statalist (http://www.stata.com/support/faqs/stat/ivr_faq.html)
>
> Then I analyze the indirect effect of the IV on the DV through the MV using nlcom on the previously stored coefficients (http://www.ats.ucla.edu/stat/stata/faq/mulmediation.htm).
>
> To calculate the direct effect, according to Statalist (http://www.ats.ucla.edu/stat/stata/faq/mulmediation.htm), I just need to regress the IV with the DV. However, generally you are not allowed to use this IV (which is an instrumented variable in (1)) in (3). How I am supposed to do so?
>
> Helena Garriga
>
>
>
>
> *
> *   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/
>



-- 
Stas Kolenikov, also found at http://stas.kolenikov.name
Small print: I use this email account for mailing lists only.

*
*   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/