What Is Meta-Analysis? Pooling Effect Sizes, Heterogeneity, Publication Bias
Daniel Ha · Seoul National University PhD StudentWhat Is Meta-Analysis?
Meta-analysis quantitatively combines results across studies of the same question to estimate an overall effect size and confidence interval that is more stable than any single study. As a synthesis of evidence, it sits at the core of evidence-based research.
How Does It Differ From a Systematic Review?
A systematic review is the process of collecting and appraising studies exhaustively through a predefined protocol, while meta-analysis is the statistical synthesis of the results those studies report. Meta-analysis is normally performed on top of a systematic review.
How Do You Conduct a Meta-Analysis?
Work in order from literature collection to bias checks.
Step 1: Search and Screen the Literature
Search databases following the PRISMA flow and screen studies against inclusion and exclusion criteria.
Step 2: Compute Effect Sizes
Calculate a standardized effect size for each study (Cohen's d, correlation r, odds ratio, and so on).
Step 3: Pool the Effect Sizes
Compute the weighted average effect size and present it in a forest plot.
Step 4: Test Heterogeneity and Bias
Examine variation in results across studies (heterogeneity, I²) and check for publication bias.
What Is the Difference Between Fixed and Random Effects Models?
A fixed effect model assumes all studies share one true effect, while a random effects model assumes the true effect can differ between studies. When heterogeneity across studies is substantial, the random effects model is appropriate.
How Do You Check for Publication Bias?
Publication bias is the problem of significant results being published preferentially, which inflates the pooled effect. Check it through asymmetry in the funnel plot and Egger's test.
Summary
Meta-analysis synthesizes multiple studies through systematic collection, effect size computation, pooling, and heterogeneity and bias checks. Choose the fixed or random effects model according to heterogeneity, and never skip the publication bias check. For organizing individual papers, read it alongside the literature review guide.