Funnel Plot Generator
A funnel plot generator turns your study effect sizes and their precision into a scatter against a 95 percent confidence funnel, the standard visual check for publication bias and small-study effects in a meta-analysis. Enter your data and read the symmetry straight from the plot.
1. Choose your data
2. Enter studies
| Study | Events (T) | Total (T) | Events (C) | Total (C) | |
|---|---|---|---|---|---|
3. Plot title
Free download · no watermark, no sign-up
What the plot shows
Each point is one study, placed by its effect estimate and its standard error. Precise, larger studies sit near the top; smaller, noisier studies spread across the wide base. In the absence of publication bias the points scatter symmetrically inside the funnel around the pooled estimate of 0.66. A gap in one lower corner can signal that small studies with unwelcome results went unpublished.
How to make a funnel plot
You build a funnel plot from the same study-level data you would use to pool an effect. Each study needs an effect estimate and a measure of its precision.
- 1. Enter each study's effect. Add one row per study with its effect size, using raw counts, means, or a pre-computed estimate, just as you would for pooling.
- 2. Let the tool derive the precision. The vertical axis uses each study's standard error, which the generator computes from your data, so the most precise studies sit near the top.
- 3. Read against the confidence funnel. The shaded region is the 95 percent funnel centred on the pooled effect. Points should scatter symmetrically inside it.
- 4. Export the figure. Download a high-resolution image for your manuscript or supplement once the plot looks right.
Reading the funnel for bias
Symmetry is the signal. When the large, precise studies near the top and the smaller studies spread evenly below them all sit inside the funnel, the evidence base looks complete. When the points lean to one side, or a corner of the funnel is conspicuously empty, the published record may be missing small studies that found nothing or found harm, the pattern unpacked in the guide to detecting publication bias. Because the same effect estimates and standard errors drive both displays, it is worth building the matching forest plot of the pooled effect and reading the two together.
The funnel is a screening tool, not a verdict. Pair it with the heterogeneity output from the meta-analysis calculator, and for the wider picture of how this plot fits a synthesis see the comparison of the forest plot versus the funnel plot and the primer on how a meta-analysis is built.
Common mistakes to avoid
- Reading a funnel from too few studies. With fewer than ten studies, random scatter looks like asymmetry. The plot is unreliable as a bias check until you have a reasonable number of points.
- Equating asymmetry with publication bias. Genuine heterogeneity, varying study quality, or chance can all bend a funnel. Asymmetry is a prompt to investigate, not a conclusion.
- Plotting effect against sample size. The vertical axis should be the standard error, the proper measure of precision. Sample size alone distorts the shape, especially for binary outcomes.
- Skipping a formal test. When you have enough studies, back up the visual read with an asymmetry test rather than relying on the eye alone.
Worried your funnel looks asymmetric?
Interpreting publication bias takes judgement, not just a plot. A methodologist can run the formal asymmetry tests, weigh the alternative explanations, and write the limitations paragraph your reviewers will look for.
Talk to a methodologistFrequently asked questions
What does a funnel plot show?
A funnel plot displays each study as a point positioned by its effect estimate on the horizontal axis and its standard error on the vertical axis, with the most precise studies near the top. In the absence of publication bias the points scatter symmetrically inside an inverted funnel around the pooled effect. Asymmetry, such as a gap in one lower corner, can indicate that small studies with unfavourable results were never published.
How many studies do you need for a funnel plot?
Funnel plots and formal tests for asymmetry are generally not recommended with fewer than ten studies, because with few points it is hard to distinguish real asymmetry from random scatter. The plot will still draw with fewer studies so you can inspect the spread, but treat any apparent asymmetry cautiously until you have a reasonable number of studies.
Does funnel plot asymmetry always mean publication bias?
No. Asymmetry can arise from publication bias, but also from genuine differences in study quality, true heterogeneity where smaller studies examine different populations, or chance. A funnel plot is a visual screening tool, not proof. It should be read alongside the heterogeneity statistics and, where appropriate, a formal asymmetry test.
What is Egger's test for funnel plot asymmetry?
Egger's test is a regression-based check that puts a number on the asymmetry you see by eye. It regresses the standardised effect on its precision and tests whether the intercept differs from zero; a significant intercept signals asymmetry consistent with small-study effects. Like the plot itself, it needs a reasonable number of studies, usually at least ten, to be trustworthy, and a significant result points to a problem to investigate rather than proving publication bias on its own.
What should a funnel plot look like with no bias?
With no bias the points form a symmetric inverted funnel. The large, precise studies cluster tightly near the top around the pooled effect, and the smaller, less precise studies fan out evenly to both sides lower down, staying mostly within the confidence funnel. The key feature is balance: roughly equal numbers of small studies fall on either side of the pooled estimate, with no empty corner.