MerQuR+Statistics Glossary

Statistics Glossary → Error bars (mean ± CI plot)

Error bars (mean ± CI plot)concept

Vertical lines drawn around each group mean in a chart to show the uncertainty of the estimate or the spread of the data.

Error bars can show three different things: the standard deviation (spread of the data), the standard error (uncertainty of the estimated mean) or a confidence interval (usually 95%). Their lengths differ a lot; each arm of a 95% CI is about twice the standard error, while the standard deviation is √n times the standard error. The caption must therefore always say what the bars represent.

A common mistake is to conclude that a difference is not significant because the 95% CI bars of two groups overlap. For two independent groups the difference can still be significant at p < 0.05 when the bars overlap somewhat; roughly, if the samples are not tiny and the overlap is no more than about half an arm, the difference is usually significant. With repeated measures, group-wise bars do not reflect the uncertainty of the within-subject comparison. An error-bar chart is meant to show size and uncertainty together, alongside the statistical test rather than in place of it.

An everyday example

Tomatoes grown with three fertilisers have mean weights of 180, 195 and 210 g. The chart draws a 95% CI bar above and below each mean: 172-188, 186-204 and 201-219 g. At a glance the reader sees not only the means but also how precise each estimate is.

In MerQur

In the independent t-test and one-way ANOVA, the 'Mean ± CI' type of the 'Chart' option plots group means with t-based confidence-interval bars. In mixed-design ANOVA the 'Chart CI' checkbox adds error bars to the interaction plot only.

Analyses where this concept appears

Independent-Samples t-TestOne-Way ANOVAMixed-Design ANOVA, Split-Plot

Related concepts

Confidence level and confidence intervalInteraction plot