#SmartPLS4 Series 39 - Moderated Mediation with Multiple Indirect Paths (Model B)

TL;DR

The model tests two indirect effects to organizational performance through collaborative culture and checks whether role ambiguity moderates each indirect effect.

Briefing Cornell Notes

Briefing

Role ambiguity does not meaningfully moderate the indirect effects in a moderated mediation model with two independent variables, one mediator, and one dependent variable. The analysis tests whether role ambiguity changes the strength of two separate indirect pathways to organizational performance: (1) organizational commitment → collaborative culture → organizational performance and (2) development → collaborative culture → organizational performance. Conditional indirect effects at higher levels of role ambiguity appear significant for the development pathway in the SmartPLS output, but the key moderated-mediation test—SmartPLS’s “index of moderated mediation”—is not provided when the moderator is applied to a specific path.

SmartPLS’s workflow starts with measurement-model checks (reliability and validity), then moves to structural-model assessment (direct effects, indirect effects, and moderation), and finally evaluates conditional mediation. In this model, the path coefficients are labeled to reflect the two indirect chains and the moderation term. The moderator (role ambiguity) is modeled as an interaction affecting the dependent variable, with the relevant moderating path coefficient labeled P5. For each indirect effect, the moderated-mediation index is constructed as a product of two path coefficients: the first-stage path into the mediator multiplied by the moderating effect on the mediator-to-outcome link.

Because SmartPLS does not automatically output the index of moderated mediation for the specific moderation placement used here, the index must be computed manually. The procedure uses bootstrapping (5,000 resamples) and exports the needed path coefficients and their standard errors from SmartPLS into Excel. For the organizational commitment indirect effect, the manual index is calculated as P1a × P5, where P1a represents organizational commitment → collaborative culture. For the development indirect effect, the manual index is calculated as P1b × P5, where P1b represents development → collaborative culture.

After computing each index, the analysis derives T statistics using the index divided by its standard error, and it also calculates lower and upper confidence bounds (lower-level and upper-level intervals). Both indices come out weak, and both T statistics are insignificant. Even under a one-tailed criterion, the T values fall below the commonly used threshold of 1.645. Confidence intervals likewise do not support a reliable moderated mediation effect.

The practical takeaway is that, despite some conditional indirect effects looking “slightly significant” in SmartPLS’s conditional output for the development pathway, the formal moderated-mediation test fails for both indirect paths. Role ambiguity therefore does not provide evidence of changing the strength of either indirect effect on organizational performance in this setup. The same labeling-and-product approach can be extended to more complex models with multiple moderators, mediators, or independent variables, with results reported and optionally plotted when significant effects emerge.

Cornell Notes

The model tests whether role ambiguity moderates two indirect effects on organizational performance through collaborative culture: (1) organizational commitment → collaborative culture → organizational performance and (2) development → collaborative culture → organizational performance. SmartPLS provides conditional indirect effects, but it does not automatically report the “index of moderated mediation” for the moderation placement used here. To address this, the index is computed manually as a product of path coefficients: for the first indirect path, P1a × P5; for the second, P1b × P5, where P5 is the moderating effect involving role ambiguity. Bootstrapped standard errors and confidence intervals are then used to calculate T statistics. Both moderated-mediation indices are weak and statistically insignificant, so role ambiguity does not meaningfully moderate either indirect effect.

Why is the “index of moderated mediation” needed, and why can’t it be taken directly from SmartPLS output here?

SmartPLS’s conditional indirect effects can show whether indirect effects differ at higher or lower moderator levels, but the formal moderated-mediation test relies on the “index of moderated mediation.” In this setup, the moderator is positioned so that SmartPLS does not output that index automatically. The analysis therefore computes the index manually using the relevant path coefficients and the moderating path coefficient (P5).

How is the moderated-mediation index constructed for the organizational commitment indirect effect?

For the organizational commitment pathway, the first-stage effect into the mediator is P1a (organizational commitment → collaborative culture). The moderation effect on the mediator-to-outcome side is P5 (the interaction involving role ambiguity and collaborative culture affecting organizational performance). The manual index is P1a × P5, using exported SmartPLS path estimates and bootstrapped standard errors.

How is the moderated-mediation index constructed for the development indirect effect?

For the development pathway, the first-stage effect into the mediator is P1b (development → collaborative culture). The same moderating coefficient P5 applies because role ambiguity is tested as the moderator of the indirect effect through the mediator-to-outcome link. The manual index is P1b × P5, again using bootstrapped outputs exported from SmartPLS.

What statistical test is used after computing the index, and what threshold is referenced?

The analysis computes T statistics as (index) ÷ (standard error). It also checks lower and upper confidence bounds (lower-level and upper-level intervals). Even with a one-tailed criterion, the T statistics are below 1.645, which is treated as the decision threshold for significance in this workflow.

What conclusion follows from the index and T-statistics results?

Both moderated-mediation indices are described as very weak, and both T statistics are insignificant. Confidence intervals likewise do not support moderated mediation. The conclusion is that role ambiguity does not moderate either indirect effect—neither organizational commitment → collaborative culture → organizational performance nor development → collaborative culture → organizational performance.

Review Questions

In what way do conditional indirect effects differ from the moderated-mediation index, and why does that matter for inference?
What exact product of coefficients is used to compute the moderated-mediation index for each indirect path (organizational commitment vs. development)?
How are T statistics and confidence intervals derived from the moderated-mediation index, and what significance threshold is used?

Key Points

1
The model tests two indirect effects to organizational performance through collaborative culture and checks whether role ambiguity moderates each indirect effect.
2
SmartPLS conditional indirect effects may look significant at certain moderator levels, but the moderated-mediation conclusion depends on the index of moderated mediation.
3
When SmartPLS does not output the moderated-mediation index for the moderation placement used, the index must be calculated manually.
4
For the organizational commitment indirect path, the moderated-mediation index is computed as P1a × P5.
5
For the development indirect path, the moderated-mediation index is computed as P1b × P5.
6
T statistics are computed as index ÷ standard error, and confidence intervals are calculated using bootstrapped lower and upper bounds.
7
Both moderated-mediation indices are weak and statistically insignificant, so role ambiguity does not moderate either indirect effect in this model.

Highlights

Role ambiguity fails the formal moderated-mediation test for both indirect pathways, even when conditional indirect effects show hints of significance.

The moderated-mediation index is manually built as a product of a first-stage path into the mediator and the moderating coefficient (P5).

Using bootstrapping (5,000 resamples), the resulting T statistics fall below the one-tailed threshold of 1.645 for both indices.