Effect size for non-parametric test
Hi,
I am using Wilcoxon rank sum test for statistical group comparison of connectivity matrices. How should effect size be computed for a non-parametric test such as Wilcoxon? Is this possible in JASP?
Thank you for your help!
Comments
Dear pyrj,
This is also known as the Mann-Whitney test. You can select that, tick "Effect size" and the output table will provide the result; the table footnote will inform you this is "the rank biserial correlation".
Cheers,
E.J.
Dear E.J.
Thank you for your answer. I investigated that the rank biserial correlation is for estimating effect size between a dichotomous variable and an ordinal variable. However, my variables are the ranks of edge strengths in the connectivity matrices of two groups, and thus neither is dichotomous. Is this a violation for the rank biserial correlation?
Best,
pyrj
I don't think so -- the two groups is the dichotomy, and the ranks (of strengths) is the ordinal variable.
E.J.
Now I understand, thank you very much!