Question 784:

Asked on July 21, 2009

Tags: Mann Whitney

1

Answer:

No answer provided yet.

Files Available For Download

1. Q784_1_q784mann_whitney.xls

The Null Hypothesis is that there is no difference between population medians. The populations are the two manufacturing facilities. For a test on population medians we can use the Wilcoxon / Mann-Whitney U test. While the computations are possible by hand, they are a bit tedious, especially where there are ties. So using software is recommended. Mini-tab and MegaStat both can conduct the Mann-Whitney U test.

In the attached excel file I show the output for MegaStat. We're comparing the median of 1.5 for Lab A and 1.0 for Lab B. We get a p-value of .4731 when adjusting for ties. Since this is well above .05, this is insufficient evidence at the level of .05 for us to reject the Null Hypothesis. We therefore conclude there isn't enough evidence to say there is a difference in Lab median pixels.