- Published
- Updated
A percentage cannot be interpreted without knowing which people form its denominator. Randomised participants, people with a measurement and attendees at a visit can represent different populations. Headings and footnotes should explain these differences rather than requiring readers to infer them.
Define the question and assumptions
Define denominator rules and presentation of missing observations for each variable. If some percentages use all participants while others use available measurements, make this visible. Category totals may differ because multiple responses are possible or because of rounding; explain the applicable rule. A population label alone does not determine the denominator of every individual cell.
Hypothetical study example
In a hypothetical results table, one arm has more missing questionnaires. A high success percentage among responders is presented as though it describes the entire arm. The team displays denominators and missing counts, separately assessing how missingness affects interpretation rather than relying on the attractive percentage alone.
Design review checks
- Verify every percentage using the numerator and denominator actually applied.
- Flag denominators that change across variables, visits or categories.
- Present missing categories and multiple-response rules consistently throughout the table.
Record the decision and its limits
An accepted table lets an independent reviewer recalculate its percentages. Do not remove essential denominator information for visual simplicity. When comparing visits, explain that contributing people may change. The narrative should not present a proportion among available observations as the certain outcome for all participants; report uncertainty and missingness limitations alongside the numerical summaries.
Frequently asked questions
- Is one N heading sufficient for every cell?
- Additional explanation is needed when data availability differs across variables.
- Must percentages always sum to one hundred?
- Multiple responses and rounding can prevent this; explain the reason.