Page 1 of 1

Wilcoxon signed-rank test??

Posted: Sun Nov 15, 2015 2:34 pm
by Gerda

I have non-normal distributed data from the same people from two different measurement times.
The paired t-test is not possible because it is not normally distributed, but when I ask SPSS to do a non-parametrical test it automatically selects the Wilcoxon signed rank test. The output tells me that the median of the differences between the two measurement times is not zero. Can I really report it like that? Is the median a correct form of significance testing...?
Sorry for my bad english and for the stupid question, but I can find the answer no where else :-(.
Thank you for answering.

Spss 22.0, Windows 8

Re: Wilcoxon signed-rank test??

Posted: Tue Nov 17, 2015 11:30 am
by GerineL
yes, it is correct. Instead of comparing the mean, like with a t-test, between two timepoints, you compare the rank of people within groups. The centrality measure that accompanies that is indeed median.
look at some youtube clips about the test to get a better understanding, for instance,

Re: Wilcoxon signed-rank test??

Posted: Tue Nov 24, 2015 3:38 pm
by Gerda
Thank you!