0

I have a pre post design with three groups. I have done Mann Whitney tests to compare pre-post change between groups 1 & 2, 1 & 3, 2 & 3.

I have two questions:

  1. Am I right in doing Mann Whitney?
  2. If pre-post change is significant between say group 1 & 2, how would I know which group has brought about a greater change?
Glen_b
  • 257,508
  • 32
  • 553
  • 939
Visal
  • 1

1 Answers1

1

  1. Look like you're doing several pairwise comparisons with Mann-Whitney test. I don't know if Mann-Whitney is appropriate because I have no idea about your study. For example, the test isn't appropriate if your groups are paired. But if you're interested in the difference between distribution and you have checked the statistical assumptions, Mann-Whitney is a good candidate.

  2. I believe the test only tells you if your distribution is different, but that doesn't tell you anything like shift in median. @Harvey has a nice picture in Why is the Mann–Whitney U test significant when the medians are equal?. If you have a significant p-value, you should check the distribution for the two groups visually.

EDIT: You might also consider multiple comparison adjustment if you care about inflation in Type 1 errors.

SmallChess
  • 6,764
  • 4
  • 27
  • 48