24. SEMinR Lecture Series. Simple Mediation Analysis

Mediation analysis in SEMinR hinges on a simple causal chain: changes in an exogenous construct affect a mediator, which then drives changes in an endogenous construct. In the PLS path model framing used here, the indirect effect is computed as the product of two path coefficients—P1 (independent → mediator) and P2 (mediator → dependent)—and the total effect combines the direct and indirect components. The practical payoff is deciding whether the relationship between independent and dependent variables runs partly (or fully) through the mediator, and whether that mediation is complementary or competitive.

Before any mediation test, the workflow demands that measurement and structural quality criteria are satisfied. For reflective mediator constructs, reliability and validity matter because weak measurement can shrink estimated indirect effects, making mediation look smaller than it truly is. The same discipline extends to the structural model: collinearity must be checked because high collinearity can bias path coefficients, sometimes turning a direct effect insignificant even when mediation exists (including cases like complementary mediation). Discriminant validity problems can also distort indirect effects, producing misleading conclusions about whether mediation exists and what type it is.

Once measurement reliability/validity and structural assumptions are in place, the mediation procedure follows a significance logic tied to P1, P2, and P3. First, the indirect effect is tested by bootstrapping the model and checking whether the relevant indirect effect is statistically significant. With one mediator, the “total indirect effect” and the “specific indirect effect” coincide; with multiple mediators, specific indirect effects must be evaluated separately.

In the worked example, the model tests whether “collaborative culture” mediates between “vision” and “organizational performance.” The analysis begins by estimating the SEMinR model (vision → collaborative culture → organizational performance) and running bootstrapping to obtain confidence intervals and significance for the indirect path. The indirect effect from vision to organizational performance through collaborative culture is found significant (the confidence interval excludes zero), supporting the claim that collaborative culture mediates the vision–organizational performance link.

The next step determines mediation type by checking whether the direct effect (P3: vision → organizational performance) remains significant. If both the direct effect and the indirect effect are significant, the result is partial mediation; if the direct effect is insignificant while the indirect effect is significant, it becomes full mediation. In this illustration, the direct path is significant, so the mediation is partial.

Finally, the mediation is classified as complementary or competitive by examining the sign relationship among paths. The product of the relevant coefficients (P1 × P2 × P3) indicates whether the mediator’s effect aligns with (complementary) or opposes (competitive) the direct effect. Here, the product is positive, indicating complementary mediation. The session closes with a reminder that the same logic generalizes to more complex models, where multiple mediators require specific indirect-effect functions and more elaborate interpretation.