Paired t-Test
Tests whether the mean difference of paired measurements equals a hypothesized value.
Description
The paired t-test compares two measurements on the same units (e.g. before/after, or two measurement systems on the same part). It forms the difference d = x₁ − x₂ for each pair and runs a one-sample t-test on these differences against μ_d₀ (usually 0). Because every unit is its own control, the between-unit variation drops out – on paired data the test is considerably more powerful than the two-sample t-test. Requirement: the differences are approximately normal; otherwise the Wilcoxon signed-rank test on the differences is the alternative.
Formulas
Assumptions
- Both values of a pair belong to the same unit
- The differences are approximately normally distributed (or n sufficiently large, CLT)
- The pairs are independent of each other
- Continuous data
Limitations
- Pairing is by row – a shifted row silently pairs the wrong values
- Sensitive to outliers in the differences
References
- Student (W. S. Gosset) (1908), The probable error of a mean, Biometrika, 6(1), 1–25
- Montgomery, D. C. & Runger, G. C. (2018), Applied Statistics and Probability for Engineers, 7th Ed., Section 10.4