Part of the Math & Statistics suite · 235 calculators

Welch's T-Test Calculator

Compare two group means without assuming equal variances — the test that should be the default for two independent samples.

Welch's t uses each group's own variance: t = (x̄₁ − x̄₂) ÷ √(s₁²/n₁ + s₂²/n₂), with the Welch–Satterthwaite degrees of freedom, which are usually fractional.

Results update as you type
Results
t statistic
1.583251
p-value (two-tailed)
Degrees of freedom
Decision
Difference in means
Standard error of the difference
CI for the difference — lower
CI for the difference — upper
Cohen's d
Reviewed September 2026. Pure mathematics: the result does not depend on where you are. Terminology follows US usage (math, parentheses, decimal point).
No account required · Google Analytics off unless allowedCalculator arithmetic runs in your browserResults update as you type
All calculations run 100% in your browser. The calculator code does not submit your figures to GlobalCalc to obtain a result.
About welch's t-test

How the welch's t-test calculator works

Welch's t uses each group's own variance: t = (x̄₁ − x̄₂) ÷ √(s₁²/n₁ + s₂²/n₂), with the Welch–Satterthwaite degrees of freedom, which are usually fractional.

It costs almost nothing when the variances really are equal and saves you when they are not, which is why modern practice recommends it over the pooled Student's t-test as the routine choice.

Formula: t = (x̄₁ − x̄₂) / √(s₁²/n₁ + s₂²/n₂)

Worked examples

Inputst statisticNote
Means 52 and 47, unequal spread1.583251t = 1.57, p = 0.126
Identical groups0t = 0, p = 1
A large clear difference4.749754strongly significant

Frequently asked questions

Why use Welch rather than Student?

Because it does not assume the two groups share a variance, and it loses almost nothing when they do. Statisticians increasingly recommend it as the default.

Why are the degrees of freedom fractional?

The Welch–Satterthwaite approximation interpolates between the two groups' information. A non-integer df is normal and correct.

What if my group sizes are very different?

That is exactly where pooling goes wrong. The pooled test can be badly anti-conservative when the smaller group has the larger variance.

Do I need to test for equal variances first?

No — and doing so is worse. A preliminary variance test inflates the overall error rate. Just use Welch.

What does the confidence interval add?

The size of the difference and its precision. A p-value alone tells you only whether zero is plausible.

Where these figures come from

Last checked: September 2026. Formulas are fixed by mathematics and do not change with tax years or regulations.