Lecture 7: Mediation (Alt 3) Flashcards
(27 cards)
What is mediation analysis used to understand?
The mechanisms or processes through which an independent variable (IV) affects a dependent variable (DV) via an intermediary, or mediator, variable.
What does mediation analysis distinguish between?
Direct effects (IV to DV) and indirect effects (IV to mediator to DV).
What does it indicate if the inclusion of the mediator renders the direct relationship non-significant?
Full mediation.
What does it indicate if the direct effect remains but is reduced when the mediator is included?
Partial mediation.
Who developed the causal steps approach to mediation analysis?
Baron and Kenny (1986).
What are the four steps of the causal steps approach?
- Show a significant total effect of the IV on the DV (Path C).
- Show the IV significantly predicts the mediator (Path A).
- Show the mediator significantly predicts the DV (Path B).
- Show that the direct effect of the IV on the DV becomes smaller or non-significant when the mediator is included (Path C′).
How is the indirect effect calculated in the causal steps approach?
A × B or C − C′.
What test is traditionally used to assess the significance of the indirect effect in the causal steps approach? What are the attributes of the test?
The Sobel test:
* strength of indirect effect divided by its standard error
* assumes a normal distribution.
What are the major limitations of the causal steps approach?
- Low statistical power requiring very large samples.
- Assumes a significant total effect (Path C), which is not necessary.
- Assumes normal distribution of the indirect effect, which is often violated (e.g., skewed or kurtotic).
What is a suppression effect in mediation analysis?
A situation where direct and indirect effects have opposite signs and cancel out, masking a significant Path C.
What is the modern and preferred alternative for testing mediation?
Bootstrapping.
What kind of technique is bootstrapping?
A non-parametric resampling technique that estimates the sampling distribution of the indirect effect directly from the data.
How does bootstrapping estimate the sampling distribution of the indirect effect?
- by repeatedly resampling the dataset with replacement (e.g., 5000 times)
- and recalculating the indirect effect for each resample
- gives an estimate of the variability of the indirect effect.
What is used to assess the statistical significance of the indirect effect in bootstrapping?
Bias-corrected 95% confidence intervals (e.g, range of plausible estimates of the indirect effect)
What does it indicate if the bootstrapped confidence interval does not include zero?
The indirect effect is considered statistically significant.
What are the advantages of bootstrapping over traditional methods?
- Requires fewer statistical assumptions (e.g., no assumption of normality)
- Has higher statistical power.
- Can detect small effects with as few as 462 participants and medium effects with as few as 71 participants.
Which SPSS tool is used to implement bootstrapping for mediation?
Andrew Hayes’ PROCESS macro.
What types of variables does the PROCESS macro allow?
Continuous and categorical predictors and outcomes, and continuous mediators.
What statistical outputs does the PROCESS macro provide?
- All relevant regression path statistics,
- Model fit indices,
- Confidence intervals for indirect effects,
- Optional Sobel tests.
What types of multiple mediator models does the PROCESS macro support?
- parallel (independent paths)
- serial (linked chain of mediators).
What can the PROCESS macro estimate in models with multiple mediators?
- The total indirect effect,
- Individual indirect paths,
- Contrasts between indirect effects.
Why is interpretive caution critical in mediation analysis?
Because mediation models often include unidirectional arrows, but when based on cross-sectional data, they cannot establish causality.
What is required for making causal claims in mediation analysis?
Strong theoretical justification or multiple studies (e.g., experiments) that establish causal links between each pair of variables.
What can challenge directional assumptions in mediation models?
- Alternative causal explanations
- third-variable confounds.
