Upload a CSV with one row per study — label, effect size, and standard error (or its confidence interval) — and get the forest plot, fixed-effect and random-effects pooled estimates, I-squared and tau-squared, a prediction interval, subgroup pooling, and a funnel plot with Egger's test. Free.
Free analyses run on up to 10,000 rows. Larger files are randomly sampled to that size — sign up to analyze your full dataset.
Pooling effect sizes and testing heterogeneity...
Sent to — Forest plot, fixed-effect and random-effects pooled estimates, heterogeneity diagnostics with a prediction interval, subgroup pooling, funnel plot with Egger's test, R code, and AI insights.
Analyze another fileStudies are pooled by inverse-variance weighting. The fixed-effect estimate weights each study by 1 divided by its variance; Cochran's Q measures the weighted dispersion around it and is tested against a chi-square distribution, giving I-squared as the share of variation beyond sampling error and the DerSimonian-Laird moment estimate of tau-squared, the between-study variance. The random-effects estimate re-weights by 1 divided by (variance plus tau-squared), and a t-based prediction interval gives the range for a new study. When a grouping column is mapped, each subgroup is pooled separately and compared with a between-subgroup Q test. Funnel asymmetry is tested with Egger's regression of each study's standardised effect on its precision, whose intercept estimates the small-study effect. Where a standard error is missing, it is recovered from the 95% interval as half the width divided by 1.96. Every formula is closed form in base R.
Use it whenever you have several studies reporting the same kind of effect and you want the combined estimate plus an honest read on how much they disagree — systematic reviews, replication sets, multi-site trials, or an internal set of experiments run on the same question.
Not for raw participant-level data (analyse that directly rather than pooling summaries), not for untransformed ratio measures, not for a single study's subgroups, and not as a way to rescue a conclusion from a literature you already suspect is selectively reported.
Built for: Researchers, evidence-synthesis and HEOR analysts, clinical and academic reviewers, and experimentation teams combining results across repeated tests
Typical data source: A spreadsheet with one row per study: the study name, its effect estimate, and either the standard error or the reported 95% confidence interval
One row per study, with the effect estimate and its precision. For example, a set of trials reporting a standardised mean difference:
Minimum 3 rows · Best with 8-60 studies
Standard-library analysis: pool effect sizes across studies. Map one row per study — a study label, the effect estimate, and either its standard error or its 95% confidence interval — and get the forest plot, the fixed-effect (inverse-variance) and random-effects (DerSimonian-Laird) pooled estimates side by side, the heterogeneity diagnostics that decide which of the two means anything (Cochran's Q, I-squared, tau-squared, plus a prediction interval for the next study), optional subgroup pooling with a between-subgroup Q test, and a funnel plot with Egger's regression for small-study effects — reported with an explicit statement of how little power that test has at your study count.
Every study's effect and 95% interval on one axis with both pooled estimates at the bottom — the whole evidence base in one picture.
The fixed-effect and random-effects estimates side by side, with which one the heterogeneity actually supports.
Cochran's Q, I-squared, tau-squared and the prediction interval — the statistics that decide how the pooled number should be read.
Each subgroup pooled separately with a between-subgroup test, so a single average cannot hide a real split.
Effect against precision — the shape selective publication distorts.
Egger's regression for funnel asymmetry, reported with how much power the test actually had at your study count.
Every formula in full, plus the log-transform trap and the limits of what pooling can fix.
Plain-English interpretation — what the numbers mean, what's significant, and what to do next.
What do all these studies say together?
Map the study label, the effect estimate, and its standard error (or its confidence interval, which the analysis inverts for you). You get the forest plot, both pooled models, and the heterogeneity statistics that tell you whether combining them into one number was a reasonable thing to do.
See our FAQ for details on pricing, data privacy, and how the analysis works. Every report includes a Methodology section showing the statistical test, assumptions checked, and diagnostics run.
Run any analysis on your own data — validated R analyses, interactive reports, AI insights, and PDF export.
Try Free — No Credit CardTell us what went wrong, in your own words. We capture the page you're on automatically, so no need to describe where you are.