By ToolzHive · Updated
An effect size describes a contrast or association. Consider its scale, uncertainty and study design; statistical significance alone does not determine practical importance.
Reference table
| Measure | Definition | Interpretation |
|---|---|---|
| Mean difference | mean₁ − mean₀ | Original outcome units |
| Cohen’s d, independent groups | (mean₁ − mean₀) / pooled SD | Standardized difference |
| Hedges’ g | J × d | Small-sample correction; J depends on degrees of freedom |
| Glass’s Δ | (mean₁ − mean₀) / control SD | Control-group variability |
| Risk difference | risk₁ − risk₀ | Absolute probability contrast |
| Risk ratio | risk₁ / risk₀ | Relative probability contrast |
| Odds ratio | odds₁ / odds₀ | Not generally equal to RR |
| Correlation r | cov(X,Y) / (SDₓ SDᵧ) | Linear association between −1 and 1 |
Worked example
Means of 60 and 50 with pooled SD 20 give d = 0.5. This is a standardized contrast, not a universal claim of practical usefulness.
Comparison pitfalls
Keep effect direction consistent. Paired and independent-group effects may have different denominators. OR and RR can differ substantially for common outcomes.
Try the related tools
- Effect Size Calculator — Cohen's d, Hedges' g, Glass's Δ, r, eta-squared and effect size conversions.
- Meta-analysis Effect Converter — Convert ratios, log effects, CIs, p-values, medians and SMDs for meta-analysis.