Estimating Mediation Effects using PLS - Can('t) you?

[Tobias Gress]

Dear PLS users:

I saw that SmartPLS users talk about and test for mediation effects in their models.

The following search on EBSCO Host did not return one article:
"marketing" and "partial least squares" and "latent variable" and "mediation".

Is there any source in a top tier journal that I could cite testing for mediation or suppression using PLS?

Regards
Tobias

[Diogenes]

Hi Tobias,
the references are the same that we had used in regression analysis.
Baron & Kenny (1986) is the more quoted, but you could see:

MACKINNON; D. P.; LOCKWOOD, C. M.; HOFFMAN, J.M.; WEST, S.G.; SHEETS, V. (2002). A Comparison of Methods to Test Mediation and Other Intervening Variable Effects. Psychological Methods, Vol. 7, No. 1, 83–104

MULLER, D.; JUDD, C.M.; YZERBYT, V. Y. (2005). When Moderation Is Mediated and Mediation Is Moderated. Journal of Personality and Social Psychology, Vol. 89, No. 6, 852–863

SHROUT, P.E.; BOLGER, N. (2002). Mediation in Experimental and Nonexperimental Studies: New Procedures and Recommendations. Psychological Methods, Vol. 7, No. 4, 422–445

Best regards,
Bido

[Tobias Gress]

Dear Professor Bido:

Thank you for the references. However, I was more looking for actual proof that studies testing for mediation in PLS got accepted in marketing or IB journals. A method can be correct and adequate for a study, however if it does not get applied and accepted/published by well known journals, examiners might question it. Even when methods get transferred form one disciplin to another, scholars might have their reservations.

There is no doubt that there is a number of studies using PLS and interaction effects published in top tier journals but what about PLS and mediation?

Kind regards
Tobias

[Diogenes]

Hi Tobias,
looking for "Mediation and PLS" in Proquest (complete document) it was found 12 papers. I read 3 of them to confirm what you want, they are:


GOLDBERG, C.B.; WALDMAN,D.A. Modeling employee absenteeism: Testing alternative measures and mediated effects based on job satisfaction. Journal of Organizational Behavior. Chichester: Sep 2000. Vol. 21, Num. 6; pg. 665 [See table 2 – p.672.]

NEUFELD, D.J.; DONG, L.; HIGGINS, C. Charismatic leadership and user acceptance of information technology. European Journal of Information Systems. Basingstoke: Aug 2007. Vol. 16, Num. 4; pág. 494. [See p. 495 and 503-505 Sobel’s (1982) test]

SHAMIR, B.; ZAKAY, E.; BREININ, E.; POPPER, M. Correlates of charismatic leader behavior in military units: Subordinates' attitudes, unit characteristics, and… Academy of Management Journal; Aug 1998; 41, 4. [p.398 and 403 they used the Baron & Kenny recommendations.]

Best regards.

Bido

[elianacarraca]

Hi Bido!

We recently received a review from one of our articles recommending that we should describe better our mediation analysis, and stating that there are "several recent advances in testing mediation since the work of Baron and Kenny (1986) that are particularly relevant for multiple mediation models [see Mackinnon, (2000, 2002); Preacher and Hayes, (2008); MacKinnon, (2009), among others]".

I agree with all this, but we tested mediation using smartPLS (already described earlier in the methodology) and, thus, bootstrapping procedures.

According to what I have read in come of the references cited above, bootstrapping is the best option available to test mediation, even superior to the Sobel test.

My doubts:

So, what I really want to know is how to clearly explain how smartPLS runs the bootstrapping... which procedure does it use?....

I read in a recent article that SEM softwares (such as LISREL) do not have bootstrapping procedures incorporated... so the authors of that article decided to test mediations using Preacher & Hayes macro apart from LISREL. Do I also need to do this or the bootstrapping procedures from smartPLS are enough? On that study the authors talk about bias corrected intervals... does smartPLS give us that (I don't think so)... and can we calculate it?

And to finish... I believe it is a more simple question... when you test for mediating effects in smartPLS, how do you do it? I usually keep the following LVs: the IV, M, and DV... the remaining LVs from the model I remove... and then I test that simple mediation: model 1 - including the paths: IV ->M; M->DV; IV->DV, and then model 2 - only including the path IV->DV. I take note of all the path values from these 2 models... total, indirect, and direct effects and follow Baron and Kenny's guidelines. Then I calculate the significance of the indirect effect apart...since PLS does not give us that.

What do you think?

Thanks a lot,

Eliana

Hi Tobias,
the references are the same that we had used in regression analysis.
Baron & Kenny (1986) is the more quoted, but you could see:

MACKINNON; D. P.; LOCKWOOD, C. M.; HOFFMAN, J.M.; WEST, S.G.; SHEETS, V. (2002). A Comparison of Methods to Test Mediation and Other Intervening Variable Effects. Psychological Methods, Vol. 7, No. 1, 83–104

MULLER, D.; JUDD, C.M.; YZERBYT, V. Y. (2005). When Moderation Is Mediated and Mediation Is Moderated. Journal of Personality and Social Psychology, Vol. 89, No. 6, 852–863

SHROUT, P.E.; BOLGER, N. (2002). Mediation in Experimental and Nonexperimental Studies: New Procedures and Recommendations. Psychological Methods, Vol. 7, No. 4, 422–445

Best regards,
Bido

[Tanja Hanslbauer]

Hi Eliana, hi everybody,

I am very much interested in the same topic

the authors of that article decided to test mediations using Preacher & Hayes macro apart from LISREL. Do I also need to do this or the bootstrapping procedures from smartPLS are enough?

And then one question with regards to

Then I calculate the significance of the indirect effect apart...since PLS does not give us that

How did you do that (sorry, I am quite a newbie)? I did not know how to calculate this and while researching on this I found the Preacher & Hayes macro which I used.

Thank you so much!
Tanja