How to present a null result on a poster

Absence of evidence is not evidence of absence

The American Statistical Association's statement on p values makes the point directly: a p value above a threshold does not show that an effect is absent, and a p value does not measure the size or importance of an effect. The 2019 Nature comment signed by more than 800 researchers was prompted in large part by this error, noting how often studies with non significant results are wrongly reported as showing no difference. The interval is what resolves it. A narrow interval centred near zero is informative; a wide one that spans both harm and meaningful benefit means the study could not tell.

Informative nulls and inconclusive ones

Two null results can mean opposite things. Suppose the smallest difference that would change practice is 5 points. A study with an interval from minus 2 to 3 has ruled out any effect that large, which is a useful finding that should shape decisions. A study with an interval from minus 7 to 10 has ruled out very little, because both a meaningful harm and a meaningful benefit remain compatible with its data. Both have p values above 0.05. Only the interval, read against that practical threshold, shows which kind of null result a poster is reporting. Writing that threshold into the methods before the analysis, rather than choosing it afterwards, is what makes this reading credible to a reviewer.

Power explains the difference

Whether a null result is informative usually comes down to power. Button and colleagues' Power failure analysis in Nature Reviews Neuroscience showed how studies with low power both miss real effects and produce unreliable estimates when they do find something. A study designed with 90 percent power for a clinically important difference that then finds a small, precise estimate is strong evidence against that difference. A pilot of 30 participants that finds nothing says almost nothing. Reporting the planned power and the achieved sample lets a reader judge which situation applies.

Common mistakes

Burying the primary outcome and leading with a positive secondary outcome is the most damaging, because it changes what the study appears to show. Others are calling p = 0.07 marginally significant, running extra analyses until something crosses a threshold, and choosing a vague title that avoids the result. Some posters omit the confidence interval precisely because it is wide. A final mistake is apologising for the result: a well designed study that finds no difference answers the question it asked and can prevent others from adopting something that does not work.

Designing the board around it

A null result can still carry a strong headline. State the question and the answer: pharmacist phone calls did not improve 90 day adherence. Use the key number to show the difference with its interval rather than the two group percentages alone. A chart can show both groups' results side by side, so a reader sees how similar they are. Give the power calculation one methods entry. Use the discussion to say what the result means for practice, what effect sizes remain possible, and what a future study would need to settle the question.

Where it shows up in the poster builder

The lab findings example is built around a partly null result: its results block leads with the doses where nothing happened, reporting p values of 0.71 and 0.44 at the two lowest concentrations, and its title carries the negative half of the finding. Key number blocks, on 27 layouts, have a detail line for the interval under the value. Chart figures hold categories and values without error bars, so the interval that makes a null result interpretable must be stated in the key number, the caption or the text. The generator writes from the brief, so say in it that the primary outcome showed no clear difference, or the draft may frame the board more positively than the data allow.

Questions people ask