I have data from four companies, W, X, Y and Z. For each company, three groups of 5 individuals (15 individuals total) take part in a test. Every individual takes two tests, test A and test B, with A taken first and once finished, and B after the individual has completed test A. So, I have two scores for each individual, one score for test A and another for test B for the same individual. All of the scores are on the scale of 0 to 20.
I want to compare the test scores, and see whether the scores of test A is better than the scores of test B separately for each company, regardless of the group the individual belongs to within the company.
My data are not normally distributed. So I am told to run a non-parametric test. What should I run in this case? Is it the Mann-Whitney U test? Should I compare test A and B in each company, separately? Are my data considered as paired? I am a bit lost. any thought or help will definitely be of golden value.