- Published
- Updated
A p-value alone does not describe effect size, clinical importance or the probability that a conclusion is true. State the hypothesis, analysis and multiplicity context before using the number to support a definitive scientific claim.
Define the question and assumptions
Connect the primary hypothesis with its analysis population and prespecified method. Explain selection when a subgroup or visit was chosen after examining outcomes. Small numerical differences around a threshold should not create an artificial scientific divide between otherwise similar bodies of evidence or similar estimated effects.
Hypothetical study example
In a hypothetical trial, the main analysis narrowly misses its planned threshold while a selected visit yields a smaller p-value. The team does not present that visit as a replacement primary result. It reports effect estimates, intervals and the planned or exploratory status of the analyses together.
Design review checks
- Link each p-value to the specific hypothesis and analytical method tested.
- Make the number of evaluated outcomes, times and subgroups visible.
- Show effect magnitude and uncertainty within the same results presentation.
Record the decision and its limits
Accepted reporting goes beyond significant and nonsignificant labels to explain data limitations. A large study may detect a small difference, while an important difference in a small study may remain uncertain. Do not relax interpretation after observing thresholds; document methods and justified changes so the analytical history can be reconstructed.
Frequently asked questions
- Is a p-value the probability a hypothesis is true?
- No. It concerns data under a specified hypothesis and statistical model.
- Does a small p-value imply a large clinical effect?
- No. Effect magnitude must be examined separately.