We use non-parametrics statistics when we don’t hit all the requirements for parametric stats. We would prefer not to use non-parametric statistics because they’re less powerful.
These procedures are used with either nominal or ordinal data. If we have interval or ratio data, we can transform them into ranks by just labeling them with highest score = 1, next is 2, etc.
The most common form of non-parametric test is the chi-square procedure.
Other non-parametric procedures include:
- Spearman correlation coefficient :: is . Similar to the Pearson correlation coefficient, but with ranked data.
- Mann-Whitney test :: like an independent-samples t-test when a study contains two independent samples of ordinal scores.
- Wilcoxon test :: Like a related-samples t-test, but with two related samples of ordinal scores.
- Kruskal-Wallis test :: Like a one-way between-subjects ANOVA, when the study has one factor with at least 3 conditions and each involved independent samples of ordinal scores.
- Friedman test :: one-way within-subjects one-way ANOVA, when the study has one factor with at least 3 levels and each involved related samples of ordinal scores.