Tests If Ordinal Data

Question 1

Tests

Answer 1

Non parametric within ppts (repeated measures)- wilcoxon matched pairs; Friedman test.
Non parametric between ppts (independent groups)- Mann Whitney U test; wilcoxon rank sum test; kruskal-Wallis test.

Question 2

When to use non parametric

Answer 2

If:
Data is nominal it ordinal.
Can’t assume sample is from a normal distribution.
Variances between 2 samples are different- common if group sizes are different (variance bigger in smaller groups).

Question 3

Non parametric within subjects

Answer 3

Wilcoxon matched pairs signed rank test-
When testing differences.
Ordinal data.
Related data(within).
Process included ranking differences across conditions.

Friedman- equivalent to wilcoxon.
Can be used when IV has more than 2 levels.

Question 4

Non parametric between subjects

Answer 4

Mann-Whitney U test:
When testing differences. 
Ordinal data. 
Unrelated data(independent). 
Can use even if two groups are different sizes. 

Kruskal Wallis test:
Like Mann Whitney- rank scores across all groups as a single set.

Pairwise comparisons:
Kruskal Wallis (& Friedman) tells of IV has significant effect across all groups.
Use multiple Mann Whitney (or wilcoxon) to test between conditions to identify where differences lie.