Chapter 10: Mediation Analysis with SmartPLS Software

1. Customer Satisfaction (SATIS) as a Mediator

The relationships among constructs in PLS-SEM can be complex and not always straightforward. To gain a better understanding of the role of SATIS in our model, its potential mediating effect on the linkage between REPUT and LOYAL (see Figure 92), and those between PRICE and LOYAL (see Figure 93) are examined in Susan’s research. This is accomplished by following the Preacher and Hayes (2008) procedure⁹⁷, which involves the use of bootstrapping in a 2-step procedure: (i) the significance of direct effect is first checked⁹⁸ using bootstrapping without the presence of the mediator SATIS in the model⁹⁹, and (ii) the significance of indirect effect”⁹ and associated T-Values— are then checked using the path coefficients when the mediator SATIS is included in the model¹⁹². This 2-step procedure is performed twice; first for testing the hypothesis six (H₆) and then subsequently for hypothesis seven (H₇). (see Figure 94 and 95)

2. Magnitude of Mediation

Once the significance of the indirect effect is established, the strength of the mediator can be examined through the use of total effect¹⁰³ and variance account for (VAF)—. Mediation analysis results are presented in Figure 96. It can be said that only 8.9% of REPUT’s effect on LOYAL can be explained via the SATIS mediator. Since the VAF is smaller than the 20% threshold level, SATIS is argued to have no mediating effect— on the REPUT->LOYAL linkage. However, 21.3% of PRICE’s effect on LOYAL can be explained via the SATIS mediator and the magnitude is considered to be partial. These findings lead us to reject hypothesis H₇ but accept hypothesis H₈ about SATIS’s mediator role.

Source: Ken Kwong-Kay Wong (2019), Mastering Partial Least Squares Structural Equation Modeling (Pls-Sem) with Smartpls in 38 Hours, iUniverse.