Mediation

Purpose:

To assess a third or "black-box" variable in the middle of a causal relationship. It tests a hypothetical causal chain where the independent variable (X or IV) affects a mediator variable (M), which in turn affects the dependent variable (Y or DV).

X → M (path a): Does the IV predict the mediator? 

M → Y (path b): Does the mediator predict the DV, controlling for the IV? 

X → Y, ignoring M (path c): The total effect of the IV on the DV.

X → Y, controlling for M (path c'): The direct effect of the IV on the DV after accounting for the mediator. 

The indirect effect (a × b): The portion of X's effect on Y that goes through M. The key question is whether this is significant (mediation).

Context Used:

• Testing a "third variable" after determining a causal relationship between the IV (X) and DV (Y).

Do NOT use mediation if: You have not established a plausible causal order (X causes M causes Y) based on theory, your sample size is small (generally N < 100), your variables are all measured at the same time point — cross-sectional mediation cannot confirm causation.

Jamovi Walkthrough:

1. Click "Modules" in the upper right-hand corner

2. Click "jamovi library"

3. Search for "medmod" under the "Available" tab and click "Install"

4. Click "medmod"

5. Click "Mediation"

6. Move the DV into the "Dependent Variable" box

7. Move the IV into the "Predictor" box

8. Move the third variable of interest into the "Mediator" box

9. Check "Labels" under "Estimates"

10. Check "Test statistics" under "Estimates"

11. Check "Confidence interval" under "Estimates"

12. Check "Percent mediation" under "Estimates"

13. Check "Path estimates" under "Additional output"

Output Interpretation:

p-value: The probability of detecting a meaningful relationship/difference when there is none. Looking for a small value (p < .05).

If p < .05, reject the null hypothesis. It IS a significant effect.

If p > .05, accept the null hypothesis. It is NOT a significant effect.

% mediation: Indicates how much of the relationship the effect accounts for.

If % is higher for an indirect effect, there is a higher effect of exposure (IV) on the outcome (DV) through the mediator.

If % is higher for a direct effect, there is a higher effect of exposure (IV) on the outcome (DV) absent the mediator.

If % is split between indirect and direct effect, it indicates a partial mediation.

If % is not split between indirect and direct effect, there is a full mediation (Note: this rarely happens in practice)

Types of Effects in Mediation:

Direct Effect (c'): The effect of the IV on the DV, controlling for the mediator.

Indirect Effect (a*b): The effect of the IV on the DV through the mediator (typically confirmed by bootstrapped CI).

Total Effect (c): The combined direct and indirect effects of the IV on the DV.

APA Format:

Appropriate data visualization: Path model or Table

Sample table: See below

Sample write-up:

See https://www.starcresto.com/blog/mediation-spss for a sample APA format write-up.

Or, see the image below. (Note: The analysis in the image was completed in SPSS rather than Jamovi, but the values and interpretation remian the same, and the output format is similar enough to follow along).