In this section, I want to discuss Nonparametric tests. Now we've seen a lot of tests up until now, we've seen all the various t-tests for two groups. We've seen the anova for more than two groups, linear regression, all of these tests very common in the literature, but they are what we call parametric tests. And we need to use a parametric test, as all those t-tests, you need the population parameter to be normally distributed. Now, think about it this way, for instance. Imagine I have two groups of patients, group A, group B. One has got early appendicitis, the other group has got perforated appendicitis. And I take sample data for white cell count from each of those groups. Now remember, I've got my sample data for white cell count. The population in this group refers to everyone out there with early acute appendicitis. I have a sample of that population in my parameter, or my statistic here that we're talking about is my white cell count. That's my variable. On the other hand, I have perforated appendicitis. My sample data white cell count, but there's a population out there, millions with perforated appendicitis. Now, what I want to do with my t-test is, to see if my two sample sets are from different populations. How likely is is it that those are two different populations. And I do those analysis from my t-tests for instance. But that would posit affect that my white cell count data for the population out there with early appendicitis. That white cell count has a normal distribution in that population. And my white cell count in the population out there with perfect appendicitis has a normal distribution as well. If that is not so. If it is not from a normally distributed distribution pattern if I can use that word, then I cannot use a normal parametric test like a t-test. Then I have to move on to use a nonparametric t-test. And I think very often we do find in the literature that the t-test was used and if we had access to the data, we might look at the fact that it wasn't that appropriate to use a parametric test. Now there's one other place to use a nonparametric test of course, and that is when we're dealing with categorical data, with the proviso that it is an ordinal type of categorical data. You've got to be able to order it in some way, because ordering plays a big role in the use of nonparametric tests. Remember the ordinal data I have to put it in some order. If someone chose they agreed or strongly disagreed, there is some order to that. If someone gave a five star rating versus a three star rating. If someone chose nine on a pain scale versus a five on the pain scale, there is some order to that. I can use, all my nonparametric tests for that kind of analysis as well.