This question is about an issue when working with an exploratory approach in mediation. When we have three variables a, b ,c (where b is the possible mediator, a is predictor, c is the outcome) and have acceptably fit levels on both the partial and full mediation solutions. In my logic when we have partial mediation, we can not have full mediation and I feel that partial model is a more plausible model. However, I wonder whether it is the other way around. I work with AMOS and use bootstrapping. Answers or references on either position would be appreciated.
Dear Çınar Kaya,
As far as I know, there are two famous approaches for testing a mediation effect: Baron and Kenny (1986) , and Preacher and Hayes(2008). Recently Baron and Kenny approach become outdated, and Preacher and Hayes approach is recommended to be followed which considered as powerful method to detect mediation. Based on their approach (bootstrapping mediation approach), we don't have to test whether we have a full or partial mediation. However, you may refer also to Hair Book (A Primer on Partial Least Squares Structural Equation Modeling (PLS-SEM) Second Edition). Hair et.al (2017) identify three types of mediation: Complementary mediation: The indirect effect and the direct effect both are significant and point in the same direction. Competitive mediation: The indirect effect and the direct effect both are significant and point in opposite directions. Indirect-only mediation: The indirect effect is significant but not the direct effect.
Article Theorizing, Testing, and Concluding for Mediation in SCM Res...
Article Mediation Analysis in Partial Least Squares Path Modeling: H...
Article Mediation Analysis: Issues and Recommendations
Regards
Dear Çınar Kaya:
I'm very interested in this topic.
Indepently of the goodness of fit indexes, I think that if the relationship between the independent "a" and the dependent "c" variables ceases to exist (not significant), then you have full mediation. If this relationship becomes weak, but still significant, then, partial mediation. Then, you check the goodness of fit indexes of your model and validate it.
This is only my opinion. Lets see if any researcher provides more information.
Regards
Dear Çınar Kaya,
As far as I know, there are two famous approaches for testing a mediation effect: Baron and Kenny (1986) , and Preacher and Hayes(2008). Recently Baron and Kenny approach become outdated, and Preacher and Hayes approach is recommended to be followed which considered as powerful method to detect mediation. Based on their approach (bootstrapping mediation approach), we don't have to test whether we have a full or partial mediation. However, you may refer also to Hair Book (A Primer on Partial Least Squares Structural Equation Modeling (PLS-SEM) Second Edition). Hair et.al (2017) identify three types of mediation: Complementary mediation: The indirect effect and the direct effect both are significant and point in the same direction. Competitive mediation: The indirect effect and the direct effect both are significant and point in opposite directions. Indirect-only mediation: The indirect effect is significant but not the direct effect.
Article Theorizing, Testing, and Concluding for Mediation in SCM Res...
Article Mediation Analysis in Partial Least Squares Path Modeling: H...
Article Mediation Analysis: Issues and Recommendations
Regards
Dear Çınar
In the classical model we have
in the model A:
"c path" between IV and DP
and in model B with mediator:
"a path" between IV and Mediator
"b path" between Mediator and DP
"c' path" between IV and DP.
The steps described in the "causal steps approach" in Baron & Kenny (1986) are:
If c' is closer to zero compared to c, and non-significant, you can conclude that M completely mediates the relationship between X and Y.
If c' is still significant, the researcher concludes that M is only a "partial" mediator of X's influence on Y.
You can test mediation by means of Sobel test.
Take into account that the classical approach received several critiques:
Zhao, X., Lynch Jr, J. G., & Chen, Q. (2010). Reconsidering Baron and Kenny: Myths and truths about mediation analysis. Journal of consumer research, 37(2), 197-206.
Alternatives have been proposed. For example, bootstrapping (MacKinnon et al., 2004) and Monte Carlo methods (Preacher & Selig, 2012). Both methods are based on the estimation of confidence interval of the indirect effect (the ab path) .
Best,
Gianmaria
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Techniques
- Structural Equation Modeling