Parametric tests require data to be "interval" or "ratio" and normally distributed. If you apply them to ordinal scales: Outliers can heavily skew the mean.
A Mann-Whitney U test revealed that the treatment group (mean rank = 18.5) had significantly lower pain scores than the placebo group (mean rank = 32.4), U = 120.5, z = -2.87, p = .004, r = 0.32. Parametric tests require data to be "interval" or
Wilcoxon Signed-Rank Test Parametric Equivalent: Paired Samples t-test U = 120.5
Ordinal data represents a unique challenge in research because it exists in the gray area between categorical and numerical information. While the values have a clear rank order, the mathematical distance between them is unknown or inconsistent. Using parametric tests like t-tests or ANOVA on ordinal data can lead to misleading results and "false precision." z = -2.87