#SmartPLS4 Series 27 - How to Report Mediation Analysis Results

Mediation results in SmartPLS4 should be reported using a tight set of ingredients: the indirect effect (whether the IV’s influence on the DV through a mediator is significant), the total effect (IV → DV without the mediator), and the direct effect (IV → DV with the mediator included). Those three pieces determine both whether mediation exists and what type it is—partial when both direct and indirect effects are significant, and full when the indirect effect is significant but the direct effect is not. The practical takeaway is that mediation reporting isn’t just about a p-value for the indirect path; it requires showing how the mediator changes the IV–DV relationship.

In the example walkthrough, mediation is tested with one mediator first (labeled as isq) while the independent variable and dependent variable are treated as internal marketing (IV) and organizational performance (DV). The analysis begins by checking the “specific indirect effects” output for the hypothesized mediation path (e.g., internal marketing → isq → organizational performance). A significant indirect effect is treated as evidence that the mediator carries the influence from IV to DV. The transcript points to the reported statistics for this indirect effect—beta, t-statistics, and a p-value below the significance threshold—as the core evidence for the mediation claim.

Next, the “total effects” output is used to confirm that the IV affects the DV overall. Here, the total effect of internal marketing on organizational performance is reported with its coefficient, t-statistics, and p-value. Then the “direct effect” is taken from the path coefficients when the mediator is included in the model. The direct effect remains significant but is reduced relative to the total effect, which signals partial mediation rather than full mediation.

The session also ties mediation type to the sign and significance pattern of the path coefficients. When the IV → mediator path and mediator → DV path are both positive (so their product is positive), the mediation is described as “complimentary.” If one of the paths were negative, the mediation would be “competitive,” even if the indirect effect is still significant. The transcript emphasizes that mediation is partial when both direct and indirect effects are significant, and it is full when the direct effect is insignificant while the indirect effect is significant.

Finally, the reporting format is made concrete. The recommended table includes: (1) total effect coefficient, t-value, p-value; (2) direct effect coefficient, t-value, p-value (from path coefficients with the mediator in the model); and (3) indirect effect details from specific indirect effects—coefficient, standard error, t-value, p-value, and the 95% confidence interval. Confidence interval interpretation is explicit: if the interval does not include zero (e.g., the interval shown for IM → isq → OP has no zero between its bounds), mediation is supported. The same table structure can be reused for other mediation hypotheses by swapping labels and mediator paths accordingly.