Meta Regression

Meta-regression is a regression analysis in a meta-analysis that relates study-level effect sizes to study characteristics, weighting studies by precision, to explain heterogeneity between studies.

메타회귀는 메타분석에서 연구별 효과크기를 연구 특성과 관련짓고 정밀도로 가중하여 연구 간 이질성을 설명하는 회귀분석이다.

How do I explain why effect sizes differ across studies in a meta-analysis?

메타분석에서 연구마다 효과크기가 왜 다른지 어떻게 설명하나요?

Meta-regression models study effect sizes as a function of study-level moderators such as sample type, dose or year, usually with a random-effects model weighted by inverse variance. It fits meta-analyses with substantial heterogeneity and enough studies per moderator.

When to use Meta Regression

• Effect sizes and their variances from multiple studies
• Study-level moderators such as design, population or year
• Substantial heterogeneity in a random-effects meta-analysis

When to use something else

• Only pooling an overall effect; use a random-effects meta-analysis
• Individual participant data are available; use multilevel models on the pooled data

Assumptions and requirements

• Effect sizes on a common metric with known sampling variances
• Enough studies relative to the number of moderators
• Associations are observational across studies (ecological)
• Random-effects specification when residual heterogeneity exists

What the analysis reports

• Moderator coefficients with confidence intervals
• Test of moderators (QM)
• Residual heterogeneity (QE, tau-squared, I-squared)
• R-squared analog
• Bubble plot

How to report it

Meta-regression indicated that [moderator] significantly moderated the effect, b = [value], 95% CI [[value], [value]], QM([df]) = [value], p = [value].

Related methods

• Random-Effects Meta-Analysis
• Subgroup Analysis
• Multilevel Meta-Analysis

Run Meta Regression on Quorant