Start Here
Not sure which test to use? The Meta-Analysis Wizard asks a few questions and gives you a direct recommendation.
Open the Meta-Analysis Wizard →Meta-Analysis Hub
Statistically synthesizing effect sizes across multiple independent studies.
Related Calculators
Medical Research Examples
Meta-Analysis: A Cochrane review pools 8 RCTs of aspirin for secondary MI prevention: random-effects d=0.36 (95% CI 0.24-0.48), p<0.001; I²=47% (moderate heterogeneity).
Meta-Regression: Meta-analysis of 12 antihypertensive RCTs (I²=68%): meta-regression on baseline SBP shows β=0.52 (SE=0.18), z=2.89, p=0.004 -- higher baseline SBP predicts larger treatment effects.
Common Mistakes
- Defaulting to a fixed-effects model without justifying why the included studies should share one true effect.
- Ignoring high heterogeneity (I² > 50%) and reporting only the pooled effect without further exploration.
- Running a meta-regression with too few studies per moderator and treating the result as confirmatory rather than exploratory.
- Not assessing publication bias (funnel plot, Egger's test) in a review with fewer than 10 studies.
- Pooling studies that are clinically too different (population, intervention, outcome definition) to meaningfully combine.
- Reporting only the pooled p-value without the effect size, its CI, and the heterogeneity statistics.
Reporting Recommendations
Journals expect more than a p-value. For each test, report the effect estimate and its 95% confidence interval alongside the p-value:
Meta-AnalysisPooled effect size, 95% CI, p-value, I², Q-statistic with p-value, τ²
Meta-RegressionRegression coefficient (slope) for the moderator, SE, z, p-value, residual Q with p-value
Recommended Learning Order
- Confidence Interval fundamentals — understand precision before pooling multiple estimates
- Meta-Analysis — pool effect sizes across studies into one overall estimate
- Meta-Regression — most advanced: explain heterogeneity with a study-level moderator
Frequently Compared Tests
Meta-Analysis vs. Meta-Regression
Meta-analysis pools an overall effect size across studies; meta-regression additionally tests whether a study-level moderator explains variation in that effect (most meaningful when heterogeneity, I², is high).
Frequently Asked Questions
Fixed-effects or random-effects model -- which should I use?
Almost always random-effects for clinical meta-analyses, since patient populations, interventions, and follow-up periods vary across studies. Fixed-effects assumes every study estimates exactly the same true effect.
What does I² mean?
I² is the proportion of total variability across studies due to genuine heterogeneity rather than chance: 0-25% low, 25-50% moderate, 50-75% high, above 75% very high.
When should I run a meta-regression?
When I² is high (≥ 50%) and you have a specific, pre-specified hypothesis about a study-level moderator (e.g. dose, age, follow-up duration) that might explain the heterogeneity.
How many studies do I need for a meta-analysis?
There is no strict minimum, but fewer than 3-5 studies makes heterogeneity statistics and funnel-plot assessments unreliable. Meta-regression needs roughly 10+ studies per moderator.
What is a funnel plot for?
A funnel plot visualizes whether smaller studies show systematically different (usually larger) effects than larger studies, which can suggest publication bias. Egger's test formally tests for this asymmetry.
How do I report a meta-analysis result?
Report the pooled effect size with its 95% CI and p-value, the model used (fixed or random effects), I² and the Q-statistic with its p-value, and note any assessment of publication bias.