Hi,
My questions are related to the application of Zhao, Lunch and Chen's (2010) mediation approach in SmartPLS. I will first explain how I applied this method before formulating my question.
Application of Zhao et al's mediation approach
In SmartPLS I construed a complex model with three IV's, 3 DV's and one mediator as follows: http://www.freeimagehosting.net/83si3 with 3 IV's, 1 Mediator and 3 DV's.
Following the SmartPLS estimation, I wanted to implement the Zhao, Lunch and Chen (2010) mediation approach as follows:
(1) acquired the 1000 bootstrapped path coefficients from the table "path coefficients" and copy pasted them in Excel,
(2) calculated a*b for each sample,
(3) calculated sd(a*b) for the 1000 bootstrapped path coefficients,
(4) calculated T-value of a*b by (a*b)/sd(a*b) and calculated p based on this,
(5) acquired the sample mean for c from the table "total effects (Mean, ST-Dev, T-values),
(6) acquired the T Statistics (|O/STERR|) for c from the table "total effects (Mean, ST-Dev, T-values) and calculated p based on this,
(7) calculated a*b+c for the total mediated effect,
(8) determined the type of mediation (complementary, competitive, indirect-only, direct-only and no-effect)
This lead to an interesting insight regarding the mediation explanation, however when I intended to report on this, I met two challenges: (1) In the Zhao et al's mediation calculation, I found mediating relationships that did not show in the SmartPLS estimation and (2) Zhao et al do not address a calculation of the R square.
Reference: Zhao, X., Lunch Jr., J. G., & Chen, Q. (2010). Reconsidering Baron and Kenny: Myths and Truths About Mediation Analysis. Journal of Consumer Research, 37(2), 197-206. http://dukespace.lib.duke.edu/dspace/bi ... sequence=1
My questions
(1) Is it actually possible to apply Zhao's mediation approach as a part of the SmartPLS analysis?
(2) Zhao et al calculates the total mediated effect (a*b+c), however, how does this influence the SmartPLS estimated R-square of the DV's and (how) should I recalculate?
(3) Do you have suggestions on how to report the findings of both analysis?
Looking forward to hearing from you,
Dennis
Zhao et al (2010) mediation approach in SmartPLS
-
christian.nitzl
- PLS Expert User
- Posts: 248
- Joined: Sat Jul 25, 2009 1:34 pm
- Real name and title:
Hi Dennis!
Ad 1) In my opinion it is the best approach to calculate the mediation effect in SmartPLS actually.
Ad 2) In the case of a partly mediation of an effect there have to be an effect on R2 normally through the integration of an additional predictor. I could imagine there are some cases without any effect when there is a full mediation. But I cannot see why this should be problem. I don’t know any paper in my context of science who correct R2 because a mediation analysis.
Ad 3) I would report it in stepwise procedure. Report first the direct effects (c) and afterward report also the direct and the indirect effects (a and b). An example you can find in my working paper on pp. 19: http://papers.ssrn.com/sol3/papers.cfm? ... id=2127701
Best regards,
Christian
Ad 1) In my opinion it is the best approach to calculate the mediation effect in SmartPLS actually.
Ad 2) In the case of a partly mediation of an effect there have to be an effect on R2 normally through the integration of an additional predictor. I could imagine there are some cases without any effect when there is a full mediation. But I cannot see why this should be problem. I don’t know any paper in my context of science who correct R2 because a mediation analysis.
Ad 3) I would report it in stepwise procedure. Report first the direct effects (c) and afterward report also the direct and the indirect effects (a and b). An example you can find in my working paper on pp. 19: http://papers.ssrn.com/sol3/papers.cfm? ... id=2127701
Best regards,
Christian