Serial mediation
Purpose
Decompose the X–Y association into total, direct and mediator-specific indirect effects.
Data and assumptions
This release uses OLS for continuous observed variables. All equations use common complete cases and categorical coding. Bootstrap resamples entire cases. Covariates require substantive justification.
Configure this method
Specify M1 before M2. The model estimates specific indirect effects through M1, through M2 and through M1 then M2. The ordering needs temporal or theoretical justification.
Variables
| Control | Input |
|---|---|
| Outcome Y | Select one variable |
| Predictors X | Select variables |
| First mediator | Select one variable |
| Second mediator | Select one variable |
| Control Variables | Select variables |
| Treat as categorical | Select variables |
| Reference categories | Choose categorical reference levels |
Analysis settings
| Parameter | Choices | Default |
|---|---|---|
| Bootstrap replications | Set as needed | 1000 |
| Random Seed | Set as needed | 20260926 |
| Include path regression tables | Set as needed | Disabled |
Analysis parameters remain visible in the main configuration area. Advanced formatting contains only the table title and numeric decimals. Methods that calculate intervals also show the significance level; 0.05 corresponds to 95% intervals.
Read the results
Focus on percentile bootstrap intervals for each indirect effect and whether they include zero. A significant total effect is not a prerequisite. Cross-sectional associations alone cannot establish temporal order or a causal mechanism.
Workflow and exports
- Upload the dataset and choose the method from the sidebar or search.
- Select the variables above, set the design-specific options, then run the analysis.
- Inspect the paper-style result. Reconfigure from the result area, or open raw output for diagnostics and sample details.
- Rename and keep result tabs as needed. Export Word, Excel, CSV or TXT; batch export can include the selected AI interpretation. Word retains figures. AI text should be checked against the estimates.