Calculate standardized effect size between two groups using Cohen's d, Hedges' g, and Glass's delta.
Last updated: March 2026
Based on |d| = 0.453
| d Value Range | Magnitude | Practical Meaning | Percentile Overlap |
|---|---|---|---|
| |d| < 0.2 | Negligible | Barely noticeable difference | >92% |
| 0.2 ≤ |d| < 0.5 | Small | Noticeable but still small | 85–92% |
| 0.5 ≤ |d| < 0.8 | Medium | Clearly visible difference | 69–85% |
| |d| ≥ 0.8 | Large | Substantial, meaningful difference | <69% |
Cohen's d is a standardized measure of effect size that quantifies the difference between two group means in units of standard deviation. Unlike t-tests (which test whether a difference exists), Cohen's d measures how large or meaningful that difference is, independent of sample size.
Cohen's d is calculated as: d = (μ₁ - μ₂) / σ_pooled, where μ₁ and μ₂ are the group means and σ_pooled is the pooled standard deviation. Related measures include Hedges' g (a bias-corrected version) and Glass's delta (which uses only the control group's standard deviation).
Effect sizes are crucial for research interpretation. A statistically significant result (low p-value) might involve a tiny, practically meaningless effect size, while a non-significant result might have a large practical effect size due to low sample size.
Did a training program improve test scores?
P-values only tell you if a difference exists, not how meaningful it is. Effect sizes quantify the magnitude of the difference, which is crucial for practical decision-making.
Yes, Cohen's d can be negative (indicating Group 2 is higher). The sign indicates direction; the absolute value indicates magnitude. Conventionally, we interpret |d|.
Use Cohen's d when variances are equal. Use Hedges' g for small samples or when correcting bias. Use Glass's Δ when one group is a control and variances differ.
A large effect size indicates a substantial, meaningful difference. In research, this suggests strong practical significance, even if sample size or p-values might vary.
Effect size is independent of sample size. A large sample can detect small effects (low p-value), but that doesn't mean the effect is large—check Cohen's d.
Context matters. In experimental research, medium effects are often satisfactory. In clinical settings, even small effects might be meaningful; in other fields, large effects might be required.
Related Tools
Information entropy.
Species diversity.
Categorical variability.
Intraclass correlation.
Smoothed trend line.
Percentage change.