Wilcoxon rank sums are two statistics T+ and T– computed from paired replicates data.
Suppose we have two sets of pairs of measurements (xi,yi), i=1,…,N for each of N experimental units. We compute differences
| di = yi – xi; i=1,…,N |
and their absolute values:
| Di = |di|. |
Then we replace Di by ranks and split the values di into two groups according to the sign of corresponding di. The sums of ranks for positive and negative differences are named Wilcoxon rank sums T+ and T–, respectively.
Wilcoxon rank sums are used to compute the test statistic for the Wilcoxon signed ranks test