I have a dataset with 20 variables, and 81 observations from two groups. I want to test whether the two groups are different.
The observations are not normal within each group so I must do a non parametric test.
If I had just one variable, I would do a wilcoxon test (wilcox.test in R). But I have 20 variables, and I don't want to make 20 wilcox.test because I want to study the data as a whole.
I thought I could find a multivariate wilcoxon test, which I did: the function mWilcoxonTest from DepthProc package.
It is presumably easy to run and I did:
mWilcoxonTest(data1[,c(3:22)], data2[,c(3:22)])
My data1 and data2 datasets contain 7 and 74 observations respectively, and the 20 variables are found in columns 3 to 22.
But unfortunately, every time I run this command, the results are different:
W = 200, p-value = 0.3253
W = 225, p-value = 0.5733
etc.
This doesn't make any sense!!
So (1) Do you understand what is going on?
Or (2) do you know another way to do this test in R?
Thanks!