I have $N=24$ controls and $M=10$ patients. ($M$ can be $20$ later.)
Each subject has one $3D$ image, each element of the image is a discrete value (integer) ranging from $0$ to $4$ $[0,1,2,3,4]$. It's like a score.
I want to find a set of voxels whose scores are significantly different between two groups, i.e. in which regions the scores were significantly increased in patient group compared to controls.
As it is difficult to assume a parametric distribution of the scores across the whole image, I applied Wilcoxon rank sum test for each voxel, using $N$ integers distribution from controls and $M$ integers distribution from patients.
1. My first question is that if Wilcoxon rank sum test is a good choice for this kind of discrete data distributions.
2. My second question is about multiple comparison controlling false discovery rate for this test is appropriate.
If you have better idea please teach me.