Non-parametric Tests
Statistical tests that do not assume a specific population distribution, often based on ranks, used when data are ordinal, skewed, or samples are small.
Non-parametric tests make minimal assumptions about the shape of the population distribution, unlike parametric tests such as the t-test and ANOVA, which assume approximate normality. Many work by converting values to ranks. Common examples include the Mann-Whitney U test for two independent groups, a signed-rank test for paired data, a rank-based test for three or more groups, and rank correlation for association.
They suit ordinal data such as single Likert items, skewed distributions, small samples, and datasets with extreme outliers. The trade-off is somewhat lower power than their parametric equivalents when parametric assumptions actually hold.
In a thesis, non-parametric tests are your defensible fallback when assumption checks fail. State in the methodology chapter how you decided between parametric and non-parametric options, and report medians rather than means for variables analyzed by ranks.
Writing the thesis this term belongs to?
CiteDash takes a thesis from research question to a compiled document, with AI that cites only real papers and verifies every claim against its source.