44. SPSS AMOS - Moderated Mediation | Hayes Model 14 in AMOS

Moderated mediation in AMOS hinges on one question: does a moderator change the *strength* of an indirect effect, not just whether the moderator affects a direct path. In this walkthrough, the indirect effect of organizational commitment (OC) on organizational performance (OP) through collaborative culture (CC) is tested for moderation by role ambiguity (RA). The analysis treats RA as a moderator of the “B path” (the CC → OP link), meaning RA alters how strongly CC translates into OP—so the indirect effect OC → CC → OP can strengthen or weaken depending on RA.

The model is built around a standard mediation structure with an added moderation component consistent with Hayes Model 14 logic. OC predicts CC (the “a path”), CC predicts OP (the “b path”), and RA is positioned to influence the dependent variable through the moderated segment of the mediation. To implement this in AMOS, the workflow centers the relevant variables—specifically CC and RA—by subtracting their means from each composite score. Centering is used to create a clean interaction term and reduce multicollinearity in the product term.

Next comes the interaction term: the centered CC score is multiplied by the centered RA score to form the product term (the interaction). In AMOS, this interaction term is then added to the structural model so it can affect the endogenous variable on the moderated path. The model’s paths are explicitly labeled so the software can compute the needed conditional indirect effects: the a path (OC → CC), the b path (CC → OP), the direct c path (OC → OP), the d path (RA → OP), and the interaction-driven e path (the interaction term’s effect on OP).

The key step is defining “new estimates” in AMOS using the Estimate function. The indirect effect is computed as the product of the a and b paths (a*b). To probe moderated mediation, the indirect effect is evaluated at two RA levels: one standard deviation below the mean (“low RA”) and one standard deviation above the mean (“high RA”). The conditional indirect effect at low RA is calculated by combining the b path with the interaction effect scaled by the negative RA standard deviation; the high-RA version uses the positive standard deviation. This produces conditional indirect effects that reveal whether mediation holds and whether it changes across RA.

After running bootstrap-based inference (5,000 bias-corrected bootstrap samples with 95% confidence intervals), the results show significant moderation on the relevant path (the interaction term is significant), and mediation exists at both low and high RA levels. The conditional indirect effect increases as RA rises: the indirect effect at low RA is significant, and the indirect effect at high RA is significant and larger, indicating RA strengthens the indirect relationship.

Finally, the analysis tests the moderated mediation formally using the index of moderated mediation. In this setup, the index is computed as the product of the interaction-driven e path and the a path (e*a). A significant index indicates the indirect effect’s slope differs from zero across RA levels—confirming that the indirect effect OC → OP through CC is genuinely moderated by role ambiguity, not merely that separate conditional effects are significant. The walkthrough also cross-checks the AMOS results against a Hayes PROCESS Model 14 run (with centered continuous variables and defined products), finding matching values for the index of moderated mediation and related interaction effects.