How to add significance bars and asterisks to a plot

The horizontal bar spanning two groups with an asterisk above it. Simple to draw, easy to get wrong.

The convention

• * for p < .05
• ** for p < .01
• *** for p < .001
• ns or nothing for non-significant

Whatever you use, define it in the figure legend. Journals differ, so do not assume the reader shares your scheme.

Exact p-values are increasingly preferred

Many journals now prefer the actual value printed above the bar. p = .003 carries more information than ** and avoids implying that .049 and .0001 are similar findings. Below .001, write p < .001 rather than a string of zeros.

Which comparisons to show

Only the ones you planned. A plot with every possible pairwise bar signals fishing and is unreadable. If you ran a post-hoc correction, the bars must reflect the corrected p-values, and the legend should name the correction.

Layout

Stack bars at increasing heights so they do not overlap, keep the shortest comparisons lowest, and leave clear space above the tallest error bar. If you need more than three or four bars, a table may serve the reader better.

In Plotara

Run the test in the Statistics tab and send the result onto the plot as significance bars. The bars carry the corrected p-values from the test you ran, so figure and statistics cannot drift apart.