Definition

T-test

A test of whether two means differ: one sample against a known value, two independent groups, or paired measurements.

The t-test asks whether an observed mean difference is large relative to the noise (standard error). Three flavors: one-sample (mean vs a benchmark), independent-samples (two groups), and paired (same units twice — before/after, matched pairs).

Welch’s variant, default in modern software, drops the equal-variance assumption at almost no cost. With small samples and clearly non-normal data, the rank-based Mann-Whitney (independent) or Wilcoxon (paired) tests are the standard alternatives.