Which test is used for related samples with ordinal data?

Which test is used for related samples with ordinal data?

• A. The Wilcoxon (T) matched pairs signed ranks test
• B. The Mann-Whitney U test
• C. The T-test for dependent samples
• D. The Kruskal-Wallis test

Answer: (A)

For related (paired) samples with ordinal data, a nonparametric approach is most appropriate because you can’t rely on normality or equal intervals between values. The Wilcoxon matched-pairs signed-ranks test is designed exactly for this situation. It looks at the differences between paired observations, keeps the direction of each difference, and then uses the magnitudes of those differences by ranking their absolute values. By comparing the sums of ranks for positive differences and negative differences, it tests whether there is a systematic shift between the two related conditions—essentially checking if the median difference is zero. This test is robust to nonnormal data and requires only the data to be at least ordinal, which fits ordinal measurements nicely. The other options fit different scenarios: the Mann-Whitney U test compares two independent samples, not related ones; the T-test for dependent samples is the parametric counterpart that assumes interval data and normally distributed difference scores; and the Kruskal-Wallis test handles more than two independent groups.

For related (paired) samples with ordinal data, a nonparametric approach is most appropriate because you can’t rely on normality or equal intervals between values. The Wilcoxon matched-pairs signed-ranks test is designed exactly for this situation. It looks at the differences between paired observations, keeps the direction of each difference, and then uses the magnitudes of those differences by ranking their absolute values. By comparing the sums of ranks for positive differences and negative differences, it tests whether there is a systematic shift between the two related conditions—essentially checking if the median difference is zero.

This test is robust to nonnormal data and requires only the data to be at least ordinal, which fits ordinal measurements nicely. The other options fit different scenarios: the Mann-Whitney U test compares two independent samples, not related ones; the T-test for dependent samples is the parametric counterpart that assumes interval data and normally distributed difference scores; and the Kruskal-Wallis test handles more than two independent groups.