Statistics
Effect size
An effect size measures the magnitude of a relationship or difference, independent of sample size. Examples include Cohen's d, correlation coefficients, odds ratios, and mean differences. Unlike a p-value, which only signals whether an effect is detectable, an effect size tells you whether it is large enough to matter.
A finding can be highly statistically significant yet practically negligible, or clearly important yet fall short of significance in a small sample. Effect sizes separate these cases. They come in standardized forms, such as Cohen's d or Hedges' g, and unstandardized forms expressed in the original units, such as a difference of 5 mmHg in blood pressure or 3 points on a depression scale.
Reporting effect sizes with confidence intervals is standard in most reporting guidelines precisely because it moves the conversation from whether there is an effect to how big it is and how precisely it is estimated. Meta-analysis is only possible because studies report effect sizes on a common scale.
When reviewing, look for results reported as significant with no accompanying magnitude, standardized effects presented without the raw units that give them meaning, or benchmarks (small, medium, large) applied without regard to the specific field. Context determines whether a given effect is trivial or transformative.
Example
Two drugs both lowered symptoms significantly, but one had a Cohen's d of 0.15 and the other 0.8, a difference no p-value would reveal.
Relevant in
ManuscriptMind checks your manuscript for issues like this before you submit.
Get a free review