Make fewer assumptions of populations.

For data which is not normally distributed- ranks make normal distribution.

Usually have less power than parametric partners.

Cannot guarantee that non-p. test will be better than when the distibution of p. test is violated.

Unrelated t-test____Mann-Whitney U test

Related t-test____Wilcoxon matched pairs

Pearson's____Spearman Rho's

Unrelated ANOVA____Kruskal-Wallis test

Related ANOVA____Friedman test

Transform difference between two groups in ranks.

• Rank difference, ignoring +/-ve.
• Once ranks are made, make +/-ve according to sign on difference
• Calculate the sum of positive and negative ranks. The smaller of the two is called T.
• Find significance of T in a Sig. Table.

• Combine all data (from both groups), rank, then separate into groups again.
• Sum ranks for the larger group of scores (if both equal, chose either)- R¹
• Put into equation, then check sig. of U in sig. table.