Non-parametric tests

Wikis > Research > Statistics > Statistical Tests > Non-parametric tests

Use non-parametric tests when assumptions for parametric tests are not satisfied: normal distribution
homogeneity of variance
ordinal or nominal scale

Two independent sample:
• Mann-Whitney U-test – analogous to parametric t-test – does not require equal sample sizes

Three or more independent samples:
• Kruskall-Wallis one way ANOVA – analogous to one way ANOVA

Two or more correlated sample:
• Sign test
• Wilcoxon signed ranks test

Three or more correlated samples:
• Friedman Two-way analysis of variance by ranks

Cross-tabulated data:
• Fischer’s exact test

Comments are closed.