HomeStatistics & Probability › Cohen's d Effect Size Calculator

Cohen's d Effect Size Calculator

Computes the standardised difference between two group means with the pooled standard deviation (Cohen's d), the small-sample corrected Hedges' g, Glass's Δ and the common-language effect size, with Cohen's small/medium/large interpretation.

When to use

You want to express how large the difference between two groups is in standard-deviation units, e.g. for reporting alongside a t-test, power analysis or meta-analysis.

Do not use when: You need to know whether the difference is statistically significant (use t-test), the outcome is binary (use odds-ratio), or the groups are paired (d for paired designs uses the SD of the differences).

Formula

pooled_sd = √(((n_a − 1)·sd_a² + (n_b − 1)·sd_b²) / (n_a + n_b − 2)); d = (mean_a − mean_b) / pooled_sd; g = d × (1 − 3 / (4(n_a + n_b) − 9)); Δ = (mean_a − mean_b) / sd_b; CLES = Φ(d / √2)

Hedges' correction uses the Hedges (1981) approximation of the exact gamma-function factor (difference below 0.1 % for n_a + n_b ≥ 20). Cohen's benchmarks are conventions, not thresholds of practical importance.

Inputs

ParameterTypeUnitRequiredDescription
mean_anumberyesSample mean of the first (treatment) group.
sd_anumberyesSample standard deviation (n − 1) of group A. Range: > 0
n_aintegeryesSample size of group A. Range: ≥ 2
mean_bnumberyesSample mean of the second (control) group.
sd_bnumberyesSample standard deviation of group B (used alone for Glass's Δ). Range: > 0
n_bintegeryesSample size of group B. Range: ≥ 2

Outputs

OutputTypeUnitDescription
cohens_dnumber(mean_a − mean_b) / pooled SD. Positive when A is higher.
hedges_gnumberd × (1 − 3 / (4(n_a + n_b) − 9)): bias-corrected d for small samples.
glass_deltanumber(mean_a − mean_b) / sd_b: uses only the control group's SD.
pooled_sdnumber√(((n_a − 1) sd_a² + (n_b − 1) sd_b²) / (n_a + n_b − 2)).
mean_differencenumbermean_a − mean_b in the original units.
common_language_effect_sizenumberΦ(d / √2): probability that a random member of A scores higher than a random member of B (normal, equal variances).
interpretationstringCohen's benchmarks on |d|: < 0.2 negligible, 0.2 small, 0.5 medium, 0.8 large.

Example

105 vs 100, SD 15, n = 30 each: {"mean_a":105,"sd_a":15,"n_a":30,"mean_b":100,"sd_b":15,"n_b":30}{"cohens_d":0.3333,"hedges_g":0.329,"glass_delta":0.3333,"pooled_sd":15,"mean_difference":5,"common_language_effect_size":0.5932,"interpretation":"Small"}

12 (SD 2, n 20) vs 10 (SD 3, n 25): {"mean_a":12,"sd_a":2,"n_a":20,"mean_b":10,"sd_b":3,"n_b":25}{"cohens_d":0.7675,"hedges_g":0.754,"glass_delta":0.6667,"pooled_sd":2.6059,"common_language_effect_size":0.7063,"interpretation":"Medium"}

GET https://tttkmbb.com/api/v1/calculate/cohens-d?mean_a=105&sd_a=15&n_a=30&mean_b=100&sd_b=15&n_b=30

Machine access

Sources

FAQ

d or g?

They differ only by the small-sample correction; report Hedges' g when either group has fewer than about 20 observations, otherwise the two are practically identical.

When is Glass's Δ preferred?

When the intervention changes the spread of the treated group, so the control group's SD is the better yardstick.

Can d be larger than 1?

Yes. d is a ratio of the mean difference to the SD and is unbounded; d = 1 means the means are one standard deviation apart.

Related calculators