Multilevel Mediation Analysis

Multilevel mediation analysis estimates indirect effects in nested data, such as employees within teams, while separating within-cluster and between-cluster components of the mediation pathway.

다층 매개분석은 팀 안의 직원처럼 위계적으로 내재된 자료에서 매개경로의 집단 내 성분과 집단 간 성분을 구분하여 간접효과를 추정하는 방법이다.

How do I test mediation when my data are nested in groups?

자료가 집단 안에 내재되어 있을 때 매개효과를 어떻게 검정하나요?

Multilevel mediation analysis tests whether a predictor affects an outcome through a mediator when observations are nested in clusters, separating within-cluster and between-cluster indirect effects. It fits designs such as students in schools or repeated measures within persons, where ordinary mediation would understate standard errors.

When to use Multilevel Mediation Analysis

• Individual-level outcome, mediator and predictor nested in clusters such as teams or schools
• Repeated measures nested within persons
• Distinguishing within-cluster from between-cluster indirect effects

When to use something else

• Data are not nested; use simple mediation analysis
• Few clusters to estimate between-level effects; use single-level mediation with cluster fixed effects (dummies) and small-sample-corrected standard errors or a wild cluster bootstrap

Assumptions and requirements

• Correct temporal ordering and no unmeasured confounding of the mediator-outcome path
• Adequate number of clusters
• Normally distributed random effects
• Centering choices that separate within and between effects

What the analysis reports

• Within-cluster and between-cluster indirect effects
• Monte Carlo or bootstrap confidence intervals
• Direct effects and path coefficients
• Random effect variances and intraclass correlations

How to report it

The within-level indirect effect of [predictor] on [outcome] through [mediator] was significant, ab = [value], 95% CI [[value], [value]].

Related methods

• Mediation Analysis
• Multilevel Linear Model
• Multilevel SEM

Run Multilevel Mediation Analysis on Quorant