Page 1 of 1

Hayes index of moderated mediation

Posted: Thu May 26, 2016 3:15 pm
by AAljabr
Hello,

I have one question about testing moderated mediation analysis using Hayes (2015) index. The 2nd edition of Hair et al. PLS book recommended this method when testing moderated mediation relationships. However, they did not provide clear guidelines on how to do this in PLS.

I was wondering what is the best approach that I should follow when testing this type of relationship. Should I test the mediation and moderation together in the same model?, but again I have no clue about how to examine the whole moderated mediation relationship.

Or, should I use the PROCESS in SPSS to test such a relationship using Hayes index?

Best regards,
Abdulrahman

Re: Hayes index of moderated mediation

Posted: Fri May 27, 2016 7:47 pm
by cringle
Thanks. But in the book we explain how to compute the Hayes index using PLS results. The tricky part is to use the bootstrap results to compute the bootstrap outcomes of that index. Here, you need to use the SmartPLS Excel report and to manually compute things. But that's not super diffcult. A more important question is if you really want to run a moderated mediation and if you can corretly intpret the outcomes. These things are oftentimes extremely tricky. I never touched such complicated things in my own applications... Usually, the simpler the analysis, the better the results and their interpretation.

Best
Christian

Re: Hayes index of moderated mediation

Posted: Fri May 27, 2016 10:05 pm
by AAljabr
Dear Prof. Ringle,

Many thanks for your clarification.

Best regards,
Abdulrahman

Re: Hayes index of moderated mediation

Posted: Tue Aug 16, 2016 2:10 am
by 4rain
Hello, Abdulrahman. I'm as well study the issue of moderated mediation at the moment. However, I only have the 1st ed. of A Primer on Partial Least Squares, and thus have not idea how to manually compute the index. Could you please send me a e-copy of this section? I think the path coefficients can be retrieved from unstandardized path coefficient table, but CI computation still an unsolved problem.

Re: Hayes index of moderated mediation

Posted: Tue Nov 22, 2016 10:49 am
by jmbecker
Actually, you don't need to compute the results manually, if you have a moderated mediation as depicted in the book in exhibit 7.19 (i.e., where the first part of the indirect effect is moderated). In this case, the index of moderated mediation by Hayes is simply the indirect effect of the interaction term (M*Y1) on the final dependent variable (Y3), because Haye’s index of moderated mediation is only the p2*p5 from the equation in the book (based on Exhibit 7.20) and not the p1*p2 part.
p2*p5 is the indirect effect of M*Y1 on Y3. You will get this result from the SmartPLS output and also it significance from the bootstrapping without any additional calculations.

It is more complicated if the moderation happens in the second part of the indirect effect because then you have to manually calculate the index using the first part of the indirect effect p1 times the interaction effect p5 (i.e., p1*p5), which will not be provided by standard PLS outputs.

Re: Hayes index of moderated mediation

Posted: Tue Dec 05, 2017 2:42 pm
by adnanafs@gmail.ocm
Hi All

I want to know, how can I calculate moderated mediation in SmartPls3 for my thesis.

I have one IV (X) , two mediators (M1 and M2), one DV (Y) and one Moderator (W) in my model. My moderation hypotheses are second part (i.e., "the relationship between M and Y are moderated by W).

Need your support.

Regards

Muhammad Adnan Khurshid
PhD Student

Re: Hayes index of moderated mediation

Posted: Tue Dec 05, 2017 4:43 pm
by jmbecker
Lets only consider a single mediation. You can easily transfer the principle to multiple mediators using the specific indirect effects.

If your model is X --> p1 --> M --> p2 --> Y and X --> p3 --> Y and you have a moderator W which is moderating the relationship between M and Y. Therefoe you also have W*M --> p4 --> Y (i.e., p4 is the coefficient of the interaction effect)

The conditional indirect effect is in this case p1*(p2+p4*W) because p2+p4*W is the effect of M on Y accounting for a moderating effect of W.
This is equal to p1*p2+p1*p4*W. Therefore, similar to a normal moderation your moderated mediation effect is p1*p4. Hayes (2015) calls this the index of moderated mediation.
This is the one that you want to test for moderated mediation. The problem is that you need to calculate this effect on your own and then assess its significance using the Samples data form the path coefficient in the bootstrapping output to calculate p-values or confidence intervals.
In general, the principal for this is similar as in Nitzl, C., Nitzl, C., Roldan, J. L., Roldan, J. L., Cepeda, G., & Cepeda, G. (2016). Mediation analysis in partial least squares path modeling: Helping researchers discuss more sophisticated models. Industrial management & data systems, 116(9), 1849-1864.

Hayes, A. F. (2015). An index and test of linear moderated mediation. Multivariate Behavioral Research, 50(1), 1-22.

Re: Hayes index of moderated mediation

Posted: Wed Dec 06, 2017 7:38 am
by adnanafs@gmail.ocm
Dear Prof. Becker

Thank you so much for the clarification.

Adnan

Re: Hayes index of moderated mediation

Posted: Sat Apr 18, 2020 3:53 pm
by boraokur44
jmbecker wrote: Tue Dec 05, 2017 4:43 pm Lets only consider a single mediation. You can easily transfer the principle to multiple mediators using the specific indirect effects.

If your model is X --> p1 --> M --> p2 --> Y and X --> p3 --> Y and you have a moderator W which is moderating the relationship between M and Y. Therefoe you also have W*M --> p4 --> Y (i.e., p4 is the coefficient of the interaction effect)

The conditional indirect effect is in this case p1*(p2+p4*W) because p2+p4*W is the effect of M on Y accounting for a moderating effect of W.
This is equal to p1*p2+p1*p4*W. Therefore, similar to a normal moderation your moderated mediation effect is p1*p4. Hayes (2015) calls this the index of moderated mediation.
This is the one that you want to test for moderated mediation. The problem is that you need to calculate this effect on your own and then assess its significance using the Samples data form the path coefficient in the bootstrapping output to calculate p-values or confidence intervals.
In general, the principal for this is similar as in Nitzl, C., Nitzl, C., Roldan, J. L., Roldan, J. L., Cepeda, G., & Cepeda, G. (2016). Mediation analysis in partial least squares path modeling: Helping researchers discuss more sophisticated models. Industrial management & data systems, 116(9), 1849-1864.

Hayes, A. F. (2015). An index and test of linear moderated mediation. Multivariate Behavioral Research, 50(1), 1-22.
Dear Prof. Becker,
Thank you for your kind and detailed response.
As I understand, if we have two moderating variables, we will have to calculate manually. You have also emphasized that it may be enough to look at the specific indirect effects of moderating variable on dependent variable in a different topic (viewtopic.php?f=5&t=18989&p=30406&hilit ... ion#p30406). But if we have more than one moderating variables, in the new releases of SmartPLS (eg 3.2.8 or 3.2.9), is it enough to look at the specific indirect effects of the moderating effect on dependent variable? Or have to we calculate this index with manually?

Stay safe and healthy
Best regards,

Re: Hayes index of moderated mediation

Posted: Tue Apr 21, 2020 6:47 am
by boraokur44
Also, When the B model in Figure 1 in Hayes's (2015) study is applied;
A formula is suggested for Index of moderated mediation:
a*b3.

We can achieve these two values (a and b3) separately in SmartPLS. Therefore, Index of moderated mediation can be determined. But how do you get boostrapping results of this index? How can we determine if it is significant? Is it possible to find confidence intervals?

Re: Hayes index of moderated mediation

Posted: Fri Jul 08, 2022 5:32 am
by andreyandoko
I have the same problem with index of moderated mediation. I can calculate the index from Hayes formula, but I do not know how to compute the CI and p value. Can anyone help ?
Thank you


Andrey ANdoko