MerQuR+Statistics Glossary

Statistics Glossary → Funnel plot

Funnel plotconcept

A plot of each study's effect size against its precision (usually its standard error), used in meta-analysis to look visually for small-study effects such as publication bias.

The effect size is on the horizontal axis and the standard error on the vertical axis (inverted, so precise studies are at the top). Large, precise studies cluster in a narrow band around the common effect, while small studies scatter more widely; without bias the result is a symmetric, inverted funnel. If one side of the funnel is empty, especially the corner where small, non-significant studies should be, those studies may not have been published.

Asymmetry is not proof of publication bias. It can also arise because small studies genuinely involve different patients or interventions, because of lower methodological quality, heterogeneity, or chance. A contour-enhanced funnel plot shades the regions of significance at p = 0.10, 0.05 and 0.01: if the apparently missing studies fall in the non-significant region, publication bias is more likely; if they fall in the significant region, other causes are more likely. Visual reading is subjective and unreliable with fewer than 10 studies, so it is supported by statistical tests such as Egger's test.

An everyday example

Twenty studies of a dietary supplement are placed on a funnel plot. The large studies cluster around a small effect, but all the small studies lie on the right, in the region of large positive effects. The empty lower-left corner suggests that small studies with unfavourable results may not have been published.

In MerQur

In Meta-Analysis the funnel plot is part of the publication-bias assessment; the 'Publication bias: Contour funnel' option adds significance bands at p = .10/.05/.01 to the plot.

Analyses where this concept appears

Meta-Analysis

Related concepts

Publication biasEgger's and Begg's tests