T-Test Calculator

At its core, a t-test asks whether the average outcomes of two groups differ by more than chance would explain. Researchers and analysts rely on it to test hypotheses, from comparing two treatments head to head to checking whether a revised process actually improves results.

Which version applies depends on how your data is structured: an independent samples t-test for two unrelated groups, or a paired samples t-test when the same subjects are measured twice. This calculator runs the independent version from summary statistics: enter each group's mean, standard deviation and size. The t-statistic is the difference between the means divided by its standard error, √(s1²/n1 + s2²/n2), which does not assume equal variances, so the calculator pairs it with the Welch–Satterthwaite degrees of freedom, as the NIST handbook does (close to n1 + n2 − 2 when the groups have similar spreads and sizes); the resulting two-tailed p-value tells you how surprising the observed gap would be if no real difference existed. For paired data, work out each subject's difference first and test whether the mean difference is zero; this calculator does not do that step.

Valid results depend on the test's assumptions holding, including roughly normal data and, for independent samples, comparable variances between groups. Statistical significance also is not the same as importance: a tiny but real difference can produce a small p-value, so read the effect size next to it to judge whether the finding matters in practice.