Mediation Analysis using JAMOVI: Concept, Interpretation, and Reporting Mediation in Research
Based on Research With Fawad's video on YouTube. If you like this content, support the original creators by watching, liking and subscribing to their content.
A mediator explains the mechanism by which an independent variable influences a dependent variable through an indirect pathway.
Briefing
Mediation analysis is a way to test whether an independent variable affects a dependent variable through a third, mechanism-carrying variable—rather than acting only through a direct link. In this framework, the independent variable (IV) influences the mediator (MV), and the mediator then influences the dependent variable (DV), forming a causal chain. The practical payoff is interpretive: if the IV→DV relationship shrinks once the mediator is added, that pattern supports the idea that part of the IV’s impact operates through the mediator.
The transcript lays out the core logic and the conditions under which mediation is considered present. Mediation is expected when (1) the IV significantly predicts the mediator, (2) the IV significantly predicts the DV when the mediator is not in the model, (3) the mediator has a significant unique effect on the DV, and (4) the IV’s effect on the DV decreases after the mediator enters the model. The analysis distinguishes between total effect (IV on DV without the mediator), direct effect (IV on DV with the mediator included), and indirect effect (the mediated pathway, computed as the product of the IV→MV path and the MV→DV path, labeled a×b). If the indirect effect is significant, mediation is supported; if the direct effect remains significant too, the result is partial mediation. If the direct effect becomes non-significant while the indirect effect remains significant, the pattern indicates complete mediation. If the indirect effect is not significant, there is no mediation.
To make these ideas concrete, the example uses a mediation model where Corporate Social Responsibility (CSR) is the independent variable, Service Quality is the mediator, and Customer Loyalty is the dependent variable. The model’s paths are mapped as follows: a is CSR→Service Quality, b is Service Quality→Customer Loyalty, c is the total effect of CSR on Customer Loyalty, and c′ (written as “C complement” in the transcript) is the direct effect of CSR on Customer Loyalty when Service Quality is included.
Implementation in jamovi is done through the MedMod module using “simple mediation.” The transcript recommends bootstrap estimation because it avoids assuming normality; it uses 1,000 bootstrap samples and requests 95% confidence intervals, along with percent mediation and path estimates. The output reports the indirect effect (a×b), its standard error, confidence interval, z/t statistics, and p-value; it also reports the direct effect (c′) and the total effect (c).
In the reported results, the indirect effect of CSR on Customer Loyalty through Service Quality is significant (estimate 0.215; 95% CI does not cross zero; p≈0.02). The direct effect remains significant (estimate 0.619; p<0.01), indicating partial mediation. The transcript also quantifies the mediated share: about 25% of the CSR→Customer Loyalty effect passes through Service Quality. For reporting, it provides a template-style write-up: state the mediation test, report total effect, direct effect, and indirect effect with coefficients, p-values, and confidence intervals, and interpret the mediation type based on whether direct and indirect effects are both significant.
Cornell Notes
Mediation analysis tests whether an independent variable’s influence on a dependent variable runs through a mediator that represents the mechanism. The key quantities are the total effect (c: IV→DV without the mediator), the indirect effect (a×b: IV→MV→DV), and the direct effect (c′: IV→DV with the mediator included). Mediation is supported when the indirect effect is significant; partial mediation occurs when both indirect and direct effects are significant, while complete mediation occurs when only the indirect effect is significant. In the example, CSR affects Customer Loyalty partly through Service Quality, with the indirect effect significant and the direct effect still significant (partial mediation).
What makes a variable a “mediator” rather than just another predictor?
How do total effect, direct effect, and indirect effect differ in mediation?
What criteria distinguish partial mediation from complete mediation?
Why use bootstrap estimation in mediation analysis?
In the CSR → Service Quality → Customer Loyalty example, what do the reported effects imply?
Review Questions
- In mediation terms, what do the symbols a, b, c, and c′ represent, and how is the indirect effect computed?
- If the indirect effect is significant but the direct effect is not, what mediation type should be concluded?
- What information should be included when reporting mediation results from jamovi (e.g., coefficients, p-values, confidence intervals), and why does percent mediation matter?
Key Points
- 1
A mediator explains the mechanism by which an independent variable influences a dependent variable through an indirect pathway.
- 2
Mediation is supported when the indirect effect (a×b) is significant; partial mediation occurs when the direct effect (c′) also remains significant.
- 3
Total effect (c) is the IV→DV relationship without the mediator; direct effect (c′) is the IV→DV relationship after adding the mediator.
- 4
Indirect effect significance is assessed using bootstrap confidence intervals when normality is not assumed.
- 5
In the example model, CSR influences Customer Loyalty partly through Service Quality, with an indirect effect estimate of 0.215 and a remaining direct effect estimate of 0.619.
- 6
Reporting should include total effect, direct effect, indirect effect, their p-values, and 95% confidence intervals, plus an interpretation of mediation type (partial vs complete).