I have 2 overall groups A, B. In both A and B, I have 3 individuals, each scoring a certain percentage on a test. If I want to compare the scores across group A and B, is the correct process to sum all the scores in each group across subjects and then run a fisher's exact test? Or can I take the scores from each individuals and run a two-sample t-test on the two groups of 3 scores?
the answer appears to be to use the Cochran–Mantel–Haenszel test, but it seems to me that t-test would be apt here as well. http://www.biostathandbook.com/cmh.html