Rank Randomization Tests

Two of the major drawbacks of randomization tests are the amount of required computing time and the fact that the randomization has to be carried out for each particular data set. Rank randomizations tests provide a solution to this problem, while maintaining independence of distributions.

These tests are performed by first converting the scores to ranks and then computing the randomization test. The randomization, however, can be avoided by using precalculated tables for determining the level of significance. Rank randomization tests are always less powerful than randomization tests.

Some examples of rank randomization tests are:
 

• Mann-Whitney U test (comparing two means)
• McNemar sign test (comparing distributions)
• Spearman's correlation
• Kolmogorov-Smirnov test (comparing distributions)
• Kruskal Wallis test (differences among groups)