Bookmark and Share

Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.


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

Re: st: Testing for mediation with categorical mediators and complex survey data?


From   John Antonakis <[email protected]>
To   [email protected]
Subject   Re: st: Testing for mediation with categorical mediators and complex survey data?
Date   Fri, 16 Dec 2011 14:09:51 +0100

Hi:

The thing to worry about, before figuring out the the indirect effect is whether the mediator is endogenous. It is critical to understand this issue first before testing for mediation or sgmediation; these methods use OLS and won't do the trick. You need an instrumental variable estimator as discussed here in the archives:

http://www.stata.com/statalist/archive/2011-10/msg00801.html
(if this is quite new to you, make sure you see the podcast, which is linked in the message)

So bootstrap the SE of the indirect effect when the estimate is (possibly) inconsistent will not help.

As for the estimation command, check out the user-written command -cmp-; I think that it should be able to handle what you want. Ensure to use vce(robust) (then you don't have to worry about the SEs).

After you estimate the model, you can test for the significance of an indirect effect using nlcom, e.g., (where Eq. 1 is the first stage equation of x predicting the mediator m, and Eq. 2 is the second stage equation on the mediator m predicting y).

[eq1]x*[eq2]m
(you can see how to do a "manual" Sobel test in the link above too)

This is the Sobel test.

HTH,
J.

__________________________________________

Prof. John Antonakis
Faculty of Business and Economics
Department of Organizational Behavior
University of Lausanne
Internef #618
CH-1015 Lausanne-Dorigny
Switzerland
Tel ++41 (0)21 692-3438
Fax ++41 (0)21 692-3305
http://www.hec.unil.ch/people/jantonakis
Associate Editor
The Leadership Quarterly
__________________________________________


On 16.12.2011 13:53, Galinsky, Adena M. wrote:
Hello all,

I would like to test for mediation by calculating Sobel-Goodman tests for mediated effects, using the Preacher and Hayes method of bootstrapping the standard error of the mediated effects.

I understand that the standard way to do this in Stata is to use sgmediation.

However, sgmediation does not work if either the Independent Variable (IV) or Mediating Variable (MV) is categorical, nor does it work with the svy command.

For my analysis, I am using categorical IV's and MV's (the Dependent Variable (DV) is continuous) that were collected in a survey with complex design (i.e., that are appropriately analyzed using the svy commands)

(I've already looked at "How can I do mediation analysis with a categorical IV in Stata?" found here: http://www.ats.ucla.edu/stat/stata/faq/mediation_cativ.htm. This page recommended using sureg - a command that does not work with svy nor with categorical mediating variables. It thus did not answer my question since a. It only covers the case when the IV is categorical, not when the MV is categorical and b. It does not mention what to do when using complex survey data and c. It does not mention Sobel-Goodman tests (though presumably I could write a program to calculate these tests, if I could find a solution to a. and b.)

Can you recommend what I should do?

Thank you for your help!

best,
Adena

ps  By the way, it looks like Hayes has a new SPSS macro that can handle categorical IV's and MV's - has anyone converted this into a stata version? (The SPSS version is here: http://www.afhayes.com/spss-sas-and-mplus-macros-and-code.html)

*
*   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/
*
*   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 1996–2018 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   Site index