Use when
- A group difference, within-subject change, or association needs magnitude context.
- The page compares outcomes across methods, cohorts, or time points.
- Statistical significance alone would hide practical interpretation.
Science Concepts
Science concept
Effect size
Effect size describes the magnitude of a result, making a finding interpretable beyond whether a statistical threshold was crossed.
Decision summary
Use effect size to communicate how large a difference or association is in practical terms. Pair it with uncertainty and raw-unit context so readers can judge whether the result is meaningful for the method.
| Raw scale | The endpoint unit before standardization or transformation. |
|---|---|
| Effect metric | Mean difference, standardized difference, ratio, odds, slope, or other planned metric. |
| Uncertainty | Confidence interval, credible interval, or other interval estimate. |
| Interpretation limit | Magnitude does not prove mechanism or external validity. |
Do not use when
- The underlying endpoint definition is unstable or not comparable across groups.
- The sample and analysis plan cannot support the claimed precision.
Caveats
- A standardized effect can hide whether the raw-unit difference is scientifically meaningful.
- Small samples can inflate effect-size estimates and uncertainty.
- Endpoint reliability affects the apparent magnitude of a result.
Reporting checklist
- Report the effect-size metric and why it fits the endpoint.
- Report raw-unit summaries alongside standardized values when possible.
- Report interval estimates, not only point estimates.
- State transformations and model assumptions.
- Avoid interpreting magnitude outside the method and population studied.
References
Button et al. 2013. Power failure: why small sample size undermines the reliability of neuroscience.
Discusses power, effect-size inflation, and reliability risks.
https://pubmed.ncbi.nlm.nih.gov/23571845/
Sullivan and Feinn 2012. Using effect size-or why the P value is not enough.
Accessible explanation of why effect size matters in reporting.
https://pubmed.ncbi.nlm.nih.gov/22345457/
Related surfaces
Use these related surfaces to move from the scientific method question to the relevant product page, endpoint definition, analysis tool, or adjacent guide.
