Cohen’s d Calculator for an Independent t Test

A Cohen’s d calculator for an independent t test estimates the standardized difference between the means of two separate groups. Enter each group’s mean, standard deviation, and sample size to get Cohen’s d, a measure of effect size that expresses the difference in standard-deviation units.

Formula

For two independent groups, this calculator uses the pooled-standard-deviation version of Cohen’s d:

d = (M₁ − M₂) / s_p

s_p = √[((n₁ − 1)s₁² + (n₂ − 1)s₂²) / (n₁ + n₂ − 2)]

• M₁, M₂: the two group means, in the same measurement units.
• s₁, s₂: sample standard deviations for the groups, in those same units.
• n₁, n₂: the number of observations in each independent group.
• s_p: pooled standard deviation, a weighted estimate of within-group spread.
• d: standardized mean difference, reported in standard-deviation units and without measurement units.

The pooled estimate weights each group’s variance by its degrees of freedom. This is the usual independent-groups formula when a common within-group variance is a reasonable assumption. The sign depends on the order you enter the groups: a positive value means Group 1’s mean is higher; a negative value means it is lower.

Worked Example

Using the calculator’s defaults, Group 1 has mean 12, SD 4, and n = 20; Group 2 has mean 10, SD 5, and n = 20.

1. Calculate the pooled variance: [19 × 4² + 19 × 5²] / (20 + 20 − 2) = [304 + 475] / 38 = 20.5.
2. Take its square root: s_p = √20.5 ≈ 4.5277, which rounds to 4.53.
3. Subtract the means: 12 − 10 = 2.
4. Divide the mean difference by the pooled SD: d = 2 / 4.5277 ≈ 0.4417.

Result: Cohen’s d = 0.44. The difference is positive because the first group’s mean is higher. The calculator rounds displayed outputs to two decimal places; calculations use the unrounded pooled standard deviation.

When to Use This Effect Size

Use this calculation when the groups consist of different, independent participants or units and you want to describe how far apart their means are relative to their typical variation. For instance, it can summarize a difference in test scores between two separate classes, provided the data and design support comparing their means.

Cohen’s d complements, rather than replaces, the independent-samples t test. A t test addresses evidence against a null hypothesis of equal means, while d describes the magnitude and direction of the observed difference on a standardized scale. A small p-value does not automatically imply a large effect, and a large estimated d does not by itself establish statistical significance. Sample size and variability affect the uncertainty of the estimate and the test result.

How to Interpret the Result

A d of 0.44 means the group means differ by about 0.44 pooled standard deviations, in the direction indicated by the sign. It does not mean a 44% difference, and it does not say that 44% of individuals in one group outperform individuals in the other.

Values around 0.2, 0.5, and 0.8 are often used as rough small, medium, and large reference points. They are conventions, not universal thresholds. What counts as practically important depends on the outcome, field, measurement quality, and consequences of the difference. Whenever possible, interpret d alongside the original-unit mean difference and a confidence interval.

Assumptions and Common Mistakes

• Using paired data: If the same participants are measured twice, or participants are deliberately matched, the observations are not independent. A paired design calls for a paired effect-size calculation based on within-person differences, not this pooled independent-groups formula.
• Assuming equal variances without checking: The pooled formula combines group variances and is most appropriate when a common variance is plausible. If group spreads differ substantially, consider a small-sample-corrected effect size such as Hedges’ g, or a variant using a different standardizer, and state which method you report.
• Entering standard errors instead of standard deviations: The calculator needs each group’s SD. If only standard errors are reported, recover SD as SE × √n, where n is that group’s sample size.
• Mixing scales or units: Both means and SDs must use the same scale within a group comparison. Converting all values consistently (for example, inches to centimeters) leaves d unchanged; mixing units can make it meaningless.
• Rounding too early or reversing groups: Keep full precision during calculations and round only the reported result. Reversing the group order changes the sign, not the magnitude.

Alternatives and Reporting

For unequal variances or small samples, Hedges’ g applies a correction for small-sample bias to a standardized mean difference. Glass’s delta uses the standard deviation of a designated comparison group as its denominator, which may be useful when an intervention could change variability. These effect sizes answer related questions but are not interchangeable; name the statistic and its denominator when reporting it.

A clear report gives the two means and SDs, sample sizes, mean difference, effect-size definition, and (when available) a confidence interval. For example: “Group 1 scored higher than Group 2 by 2 points; pooled-SD Cohen’s d = 0.44.” The t-test result and its uncertainty can be reported separately. A calculator using summary statistics cannot check the original data for outliers, skew, dependence, or data-entry errors, so those need attention in the analysis itself.

Frequently Asked Questions

Can I calculate Cohen’s d from an independent t statistic?
Often, yes, if you also know both group sample sizes and the t test is the ordinary pooled-variance independent-samples test. The equivalent magnitude is |t| × √(1/n₁ + 1/n₂); assign the sign using the direction of the mean difference. This conversion does not apply indiscriminately to paired tests or unequal-variance tests.

Does Cohen’s d show whether the result is statistically significant?
No. It measures a standardized difference, not a p-value. Statistical significance depends on the test, sample size, variability, and chosen significance threshold.

What if the two groups have different sample sizes?
The formula allows unequal sample sizes: each group’s variance is weighted by its degrees of freedom. Enter the actual n for each group rather than an average.

Should I use Cohen’s d or Hedges’ g?
Cohen’s d is widely recognized, while Hedges’ g adjusts d for small-sample bias. The choice depends on the analysis and reporting convention; specify which one you use, especially with small samples.

Can this calculator be used for before-and-after measurements?
Not with the independent-groups formula when the same people are measured twice. Those observations are paired, and the effect size should be based on the distribution of within-person changes.

Related

• Deployment Failed: Volume Offline? Fix It Now
• Can’t Open File for Writing? Fixes That Work
• iPhone X Touch Screen Problems: Fixed or Not?

in m1 | Group 1 mean | - | 12 | 0
in s1 | Group 1 SD | - | 4 | 0
in n1 | Group 1 sample size | people | 20 | 1
in m2 | Group 2 mean | - | 10 | 0
in s2 | Group 2 SD | - | 5 | 0
in n2 | Group 2 sample size | people | 20 | 1
out d | Cohen's d | SD units | (m1 - m2) / sqrt(((n1 - 1) * s1^2 + (n2 - 1) * s2^2) / (n1 + n2 - 2)) | 2
out sp | Pooled standard deviation | - | sqrt(((n1 - 1) * s1^2 + (n2 - 1) * s2^2) / (n1 + n2 - 2)) | 2
out diff | Mean difference | - | m1 - m2 | 2