Categorical Data Analysis Using The SAS® System, 2nd Edition

7.5. Aligned Ranks Test for Randomized Complete Blocks

When the number of blocks or treatments is small, the Friedman test has relatively low power. This results from the fact that the test statistic is based on ranking the observations within each block, which provides comparisons only of the within-block responses. Thus, direct comparison of responses in different blocks is not meaningful, due to variation between blocks. If the blocks are small, there are too few comparisons to permit an effective overall comparison of the treatments. As an example, the Friedman test reduces to the sign test if there are only two treatments. This disadvantage becomes less serious as the number of treatments increases or as the number of subjects per block ...