Cohen’s d Calculator

Work out how big a difference really is. This free Cohen’s d calculator finds the standardised effect size, Hedges’ g and a 95% confidence interval from two group means, paired data, a one-sample test or a t-statistic β€” with every calculation step shown.

Cohen's d Calculator

Four ways in β€” one clear effect size out. Pick the data you have, enter the numbers, and get Cohen's d, Hedges' g and the 95% confidence interval.

Degrees of freedom: 48 (n₁ + nβ‚‚ βˆ’ 2).

Try an example:
YOUR COHEN'S D
0.333
Small effect
Hedges' g = 0.328 Β· 95% CI [βˆ’0.225, 0.892]
0.333Cohen's d
0.328Hedges' g
[βˆ’0.225, 0.892]95% confidence interval
15.000Pooled SD (sp)
63.1%Cohen's U3
86.8%Distribution overlap
Small effect. About 63.1% of group 2 falls below the group 1 mean (Cohen's U₃), and a random score from group 1 beats a random score from group 2 about 59.3% of the time.

Effect-size scale

Where your result sits against Cohen's conventions β€” 0.2 small, 0.5 medium, 0.8 large.

d = 0.333
00.20.50.81.22.02.5+
Negligible < 0.2Small 0.2 – 0.5Medium 0.5 – 0.8Large 0.8 – 1.2Very large 1.2 – 2.0Huge β‰₯ 2.0

Calculation steps

StepWorking

Cohen’s d is a standardised difference between two means: the mean difference divided by a standard deviation, so effects can be compared across studies. A d of 0.2 is small, 0.5 is medium and 0.8 is large. The calculator above also reports Hedges’ g, which corrects the small-sample bias in d, and a 95% confidence interval for the effect.

What is Cohen’s d?

Cohen’s d is an effect size: a standardised measure of how far apart two means are, expressed in standard-deviation units rather than in the original units of measurement. It was introduced by the psychologist Jacob Cohen in 1969 and is now the standard way to report the size of a difference in psychology, education, medicine and the social sciences.

Because d is standardised, a d of 0.5 means the same thing β€” a difference of half a standard deviation β€” whether the outcome was measured in test points, reaction times or blood pressure. That makes d useful for comparing results across studies and for combining them in a meta-analysis. If you are new to the idea, read what effect size means before you go further.

A p-value tells you whether a difference is likely to be real; Cohen’s d tells you how big it is. A tiny difference can be statistically significant with a huge sample, and a large difference can miss significance with a small one β€” so always report d alongside the test result. The two numbers answer different questions.

How to calculate Cohen’s d

Which formula you need depends on the data you have. The calculator above covers all four routes:

  • Independent groups (two means): d = (x̄₁ βˆ’ xΜ„β‚‚) / sp, where sp is the pooled standard deviation of the two groups. This page uses the pooled standard deviation calculator method for sp β€” the same weighted combination of the two SDs β€” rather than duplicating its explanation here.
  • Paired data: dz = mean difference / SD of the differences. The denominator is the standard deviation of the difference scores, not the SD of either set of scores on its own.
  • One sample: d = (xΜ„ βˆ’ ΞΌβ‚€) / s, the sample mean minus the population mean, divided by the sample SD. For the test that goes with it, use our one-sample t-test calculator.
  • From a t-statistic: d = t Γ— √(1/n₁ + 1/nβ‚‚). Useful when a paper reports only the t-value and the group sizes.

The spread used in the denominator matters. If you only have a single standard deviation to hand, our standard deviation calculator finds s first.

Worked examples

Each example below is one of the calculator’s “Try an example” buttons β€” click one to see the full working.

ExampleInputsCohen’s dHedges’ g95% CI
Small effectMeans 105 vs 100, SD 15, n = 25 each0.3330.328[βˆ’0.225, 0.892]
Huge effectMeans 120 vs 100, SD 10, n = 30 each2.0001.974[1.380, 2.620]
One-sample, mediumMean 108, SD 12, n = 36, ΞΌβ‚€ = 1000.6670.652[0.306, 1.028]
From tt = 2.5, n = 25 each0.7070.696[0.136, 1.279]
Paired, mediumMean difference 5, SD of differences 10, n = 200.5000.479[0.035, 0.965]

What is a good Cohen’s d?

There is no universal “good” value β€” a good d is one that is big enough to matter in your context. Cohen’s conventional bands are a starting point, not a verdict:

|d|SizeWhat it usually means
Below 0.2NegligibleThe groups barely differ; often too small to notice.
0.2 to 0.5SmallA real but modest difference.
0.5 to 0.8MediumA difference you can usually see in the data.
0.8 to 1.2LargeA substantial difference between the groups.
1.2 to 2.0Very largeThe distributions are clearly separated.
2.0 and aboveHugeVery little overlap between the groups.

Treat the bands with care. In a tightly controlled laboratory study a d of 0.2 can be disappointing; in a field study of an expensive treatment, a d of 0.2 can be worth a fortune. What counts as “good” also depends on the costs and risks attached to the difference. And remember: a large d does not prove practical importance β€” it only says the standardised difference is large. A drug that lowers a symptom score by a huge standardised amount is still useless if the raw change is smaller than the smallest change patients can feel.

Cohen’s d vs Hedges’ g β€” which one do I report?

Report Hedges’ g whenever your samples are small, and Cohen’s d when they are large β€” or report both and let readers choose. Hedges’ g is simply d multiplied by a correction factor, J = 1 βˆ’ 3 / (4N βˆ’ 9), which removes the small-sample bias in d. With small samples d overstates the true effect, so g is the safer number; with large samples the correction is almost 1 and the two agree (for example, d = 0.333 and g = 0.328 with 25 people per group, but g = 0.479 against d = 0.500 with only 20 pairs).

Most journals accept either, and many now ask for g specifically because so many studies are small. If you are comparing your result with a published d, use d; if you are running a meta-analysis, use g. The calculator gives you both every time, so you never have to choose blind.

How do you interpret the 95% confidence interval for Cohen’s d?

The interval is the range of effect sizes that is compatible with your data. A narrow interval means the effect is estimated precisely; a wide one means it is not. Two things matter most:

  • Does it cross zero? If the interval includes 0, the data are consistent with no effect at all β€” even when d itself looks large. The first worked example above has d = 0.333 but a 95% CI of [βˆ’0.225, 0.892], so zero is inside it.
  • Is the whole interval on one side of a meaningful value? An interval of [0.4, 0.9] supports a genuinely medium-to-large effect, while [βˆ’0.1, 1.1] is too uncertain to say much at all.

The interval is in standardised units, like d itself. To see the same interval in the original units of your measurement β€” the points, seconds or millimetres β€” use the confidence interval for the difference of two means calculator instead.

Frequently asked questions

What is Cohen’s d in simple terms?

Cohen’s d is a standardised difference between two means. It tells you how many standard deviations apart the two group means are, so a d of 0.5 means the means differ by half a standard deviation. Because it is standardised, d can be compared across studies that used different scales.

What counts as a good Cohen’s d?

By Cohen’s conventions, 0.2 is small, 0.5 is medium and 0.8 is large. What counts as “good” depends on the field and the question: a small effect can still matter a great deal, and a large effect can still be unimportant in practice. Always read d together with its confidence interval.

What is the difference between Cohen’s d and Hedges’ g?

Hedges’ g is Cohen’s d multiplied by a correction factor, J = 1 βˆ’ 3 / (4N βˆ’ 9), which removes the small-sample bias in d. Report g when samples are small and d when samples are large; with big samples the two are almost identical.

Can Cohen’s d be negative?

Yes. The sign shows the direction of the difference: a negative d means the first mean is lower than the second. The size of the effect is the absolute value, so d = βˆ’0.5 and d = 0.5 are equally large.

What does a Cohen’s d of 0 mean?

A d of 0 means the two means are equal, so there is no difference between the groups at all. In practice d is rarely exactly 0; very small values are usually described as negligible (below 0.2).

How do I calculate Cohen’s d from a t-statistic?

For two independent groups of sizes n₁ and nβ‚‚, use d = t Γ— √(1/n₁ + 1/nβ‚‚). For example, t = 2.5 with 25 people in each group gives d = 0.707. Choose the “From t-statistic” tab in the calculator above and it does the working for you.

What is the 95% confidence interval for Cohen’s d?

It is the range of effect sizes that is compatible with the data, usually calculated as d Β± 1.96 Γ— its standard error. If the interval crosses zero, the data are consistent with no effect. The calculator reports this interval for every route.

Is a large Cohen’s d always practically important?

No. A large d means a big standardised difference, but practical importance also depends on the real-world size of the difference, its costs and benefits, and the context. Statistical size is not the same as practical value β€” a huge d on a trivial outcome is still trivial. Source: the pooled standard deviation and the two-sample t-test behind this calculator are defined in the NIST/SEMATECH e-Handbook of Statistical Methods, section 1.3.5.3, “Two-Sample t-Test for Equal Means”. See all 58 free statistics tools on Empirical Rule Calculators.