Effect Size and Statistical Significance
Ask how much changed
An effect size describes the magnitude of a comparison. Depending on the outcome, that could be a difference in means, a risk difference, a ratio or another measure. The numerical value needs its scale: a difference of two units has little meaning until the reader knows what the units represent.
For a hypothetical peptide assay, an increase from ten to twelve arbitrary signal units can be expressed as an absolute difference or a percentage change. Those descriptions concern the same observation but can create different impressions. Report the starting level and measurement scale where possible.
Ask what the statistical test addresses
A P value summarises how incompatible the observed data are with a specified statistical model, including its null hypothesis and assumptions. It is not the probability that a peptide works, nor a measurement of the importance of the observed difference.
The Cochrane Handbook, Chapter 15: Interpreting results and drawing conclusions advises against allowing statistical thresholds to dominate interpretation. A very precise estimate of a tiny difference may produce a small P value. A less precise estimate of a potentially substantial difference may not cross the chosen threshold. Neither result is adequately described by a yes-or-no label alone.
Read magnitude with uncertainty
The CONSORT 2025 explanation and elaboration: updated guideline for reporting randomised trials asks trial reports to present effect estimates and their precision. That combination allows readers to distinguish an estimated difference from the range of effects the analysis leaves reasonably compatible with the data. It also makes comparison across reports more informative.
Suppose two hypothetical studies estimate similar differences, but one has a much wider confidence interval. Their point estimates agree; the strength of information about magnitude differs. Calling one positive and the other negative solely because their P values fall on different sides of a threshold obscures that relationship.
Keep scientific importance contextual
For human outcomes, meaningfulness depends on the endpoint and context, including potential harms. In laboratory work, it depends on the biological question, assay behaviour and the size of variation that matters for the proposed mechanism. Statistical significance does not supply those judgements.
A useful reading entry contains the comparator, effect measure, numerical estimate, uncertainty and exact endpoint. Our research resources support that broader approach to peptide literature. If the abstract supplies only a significance claim, inspect the tables and figures before repeating it. The most informative summary explains the observed difference and how well the experiment estimated it.
Related research reading
- Confidence Intervals in Scientific Results
- Sample Size, Precision and the Strength of a Research Claim
Explore the research blog and original batch-document library for further context.