Meta-analysis
Pool compatible effect estimates while examining variation
Meta-analysis statistically combines estimates addressing sufficiently comparable questions. It usually sits within a systematic review. Common-effect and random-effects models make different assumptions about underlying effects, and random-effects modelling does not make incompatible studies comparable. Dependence among multiple effects, between-study heterogeneity and selective publication all affect the credibility of the pooled result.
Use when several studies provide compatible estimands and uncertainty information, and a pooled estimate or explanation of between-study variation would answer the research question.
Strengths
- Can increase precision across compatible studies
- Quantifies heterogeneity and supports planned moderator analysis
Limitations
- Biased studies can yield a precise biased summary
- Sparse studies give unstable heterogeneity and publication-bias assessments
Know the boundary
A pooled observational association does not become causal because it is precise.