#SmartPLS4 Series 38 - Moderated Mediation (Model B)

Moderated mediation in SmartPLS 4 hinges on one practical shift: when the moderator changes the path that feeds the mediator-to-outcome link (instead of the predictor-to-mediator link), the “index of moderated mediation” can’t be read off automatically and must be computed manually. In this setup, the moderator is role ambiguity, and the goal is to test whether role ambiguity alters the indirect effect of organizational commitment (OC) on organizational performance (OP) through collaborative culture (CC). The analysis matters because it answers a more specific question than standard mediation: not only whether OC influences OP via CC, but whether that indirect pathway strengthens or weakens depending on the level of role ambiguity.

The session starts by clarifying the model geometry. With one predictor, one mediator, one moderator, and one outcome—plus continuous variables—the moderator can be placed on different structural paths. In the earlier version, the moderator affected the relationship from X to M (the P1 path), making the interaction term feed into the mediator. Here, the moderator instead targets the relationship from the mediator to the outcome (the P2 path). That relocation changes which path coefficients combine to form the index of moderated mediation: the index becomes the product of the predictor-to-mediator path coefficient (P1) and the moderated path coefficient (P5), rather than the earlier product involving the other moderated segment.

Before any conditional mediation is assessed, SmartPLS requires the usual measurement-model checks—reliability and validity—because moderated mediation sits inside the structural-model workflow. The structural stage then proceeds in layers: direct effects and moderation effects are evaluated first, indirect effects are assessed next (to confirm mediation through CC), and only then does conditional mediation test whether the indirect effect varies across moderator levels. The conditional indirect effects are checked at low, mean, and high role ambiguity, and they come out significant across those levels—meaning the OC → CC → OP indirect effect exists regardless of moderator level.

The key question becomes whether the indirect effect changes in magnitude as role ambiguity rises. That’s where the index of moderated mediation is used: a conditional mediated effect is considered significant if the index is significantly different from zero. Because the moderator’s location differs from the earlier model, the index is computed manually. The workflow pulls bootstrapped path coefficients (including P1 and the interaction-related coefficient P5) into Excel, multiplies them to create an index for each bootstrap sample, then averages across samples to obtain an overall index. Significance is determined via confidence intervals and a t-statistic computed as index divided by its standard error.

Results show the index is significant (the confidence interval excludes zero; the t-statistic exceeds 1.645 for a one-tailed test). Interpretation follows the sign: higher role ambiguity weakens the indirect effect. The conditional indirect effects table supports this pattern—indirect effects are larger at low role ambiguity and reduced at mean and high levels. A Johnson–Neyman-style plot is then used to visualize the moderation of the indirect pathway, showing the indirect effect dropping as role ambiguity increases. Reporting ties everything together: direct effects, conditional indirect effects at low/mean/high moderator values, and the single index of moderated mediation with its confidence interval and p-value.