Hello,
I want to conduct an analysis that investigates whether there is a difference between the two groups in terms of nine dependent variables. The sample size is so small (n1= 17, n2= 9); therefore I have to run a nonparametric test. The analysis I'll use is Mann-Whitney U, I know but I stuck at one point.
If I was used Kruskal-Wallis instead of MANOVA because of assumptions of it, to understand which group get different scores from other I must used to multiple Mann-Whitney U, and I must performed correction methods (e.g., Bonferroni) for significance value.
However, now I run Mann-Whitney U as mentioned above. Should I use any correction for the current study, what do you think?
Different professors told different strategies, so I wonder what you think.
Theoretically, nine dependent variables do not correlate with each other. Five of them is intelligence scores, two variables are language and math scores which were derived from participant's report cards, and the other two variables are their teachers' evaluation about somethings.
Now again, should I use any correction for the current study? If your answer is yes, how many corrections should I use? Nine, because I have nine variables? Three, because I investigate three constructs although they had factors?
Or, should I different numbers of correction for different constructs? For instance, there must be five corrections when I analysis intelligence; however, I need two corrections to investigate language and math scores?
If your answer is no, why?
Thank you a lot!