21,418
Views
49
CrossRef citations to date
0
Altmetric
Teacher's Corner

The Wilcoxon–Mann–Whitney Procedure Fails as a Test of Medians

ORCID Icon, , &
Pages 278-286 | Received 01 Jan 2015, Published online: 15 Mar 2018

Figures & data

Figure 1. Emergency department visits in 12 months for the Puff City study.

Figure 1. Emergency department visits in 12 months for the Puff City study.

Figure 2. Changes in postoperative nausea after aromatherapy.

Figure 2. Changes in postoperative nausea after aromatherapy.

Figure 3. No difference in WMW test, but very different medians.

Figure 3. No difference in WMW test, but very different medians.

Figure 4. A difference by WMW test, but very different medians. But in wrong direction!

Figure 4. A difference by WMW test, but very different medians. But in wrong direction!

Figure 5. A global inconsistency.

Figure 5. A global inconsistency.

Figure 6. Bubble plot representing the estimate of p″ = Pr(X1<X2) + Pr(X1 = X2)/2 for toy example.

Figure 6. Bubble plot representing the estimate of p″ = Pr(X1<X2) + Pr(X1 = X2)/2 for toy example.

Figure 7. ROC curve area representing the estimate of p″ = Pr(X1<X2) + Pr(X1 = X2)/2 for toy example.

Figure 7. ROC curve area representing the estimate of p″ = Pr(X1<X2) + Pr(X1 = X2)/2 for toy example.

Figure 8. Dominance diagram representing the estimate of p″ = Pr(X1<X2) + Pr(X1 = X2)/2 for toy example.

Figure 8. Dominance diagram representing the estimate of p″ = Pr(X1<X2) + Pr(X1 = X2)/2 for toy example.
Supplemental material

Supplementary Materials

Download Zip (23.5 KB)