I have one group (n=21). And on this group i determine efficiency level(%) with the use of 3 different model specifications. So for every unit i have 3 results of efficiency. I am interested which statistics test should i use to determine if there is difference if i use different model on the same unit. There is no normal distribution of results and sample is small. I would like to find evidence that is not matter which model i use i get the same efficiency result. Thank you.