Comparing Two Cohorts, Pre and Post scores

Question

I'm working on a research project, and I could use a little assistance in picking the correct test.

There are two cohorts that took a pre-test. Subsequently, the intervention group was given a lecture on the subject while the control group did self-study. Both groups then took the identical post-test.

My current thought was that I could compare the mean change (delta) in scores for each cohort and use a T-test. We have a very small sample size: only 20. For that reason, I was thinking that a non-parametric test may be useful. I have read about Wilcoxon Rank Sum test, but I have a few questions.

1.) Is that the correct test to use?

2.) In JMP, the software I'm using, they have multiple ways to run Wilcoxon Rank Sum tests, and I can't figure out which way to do it.

a. Analyze-->Fit Y by X-->Delta Score by Cohort-->Nonparametric-->Wilxocon.

--This method gives me "2 sample test, normal approximation" and "1 way test, chiSquare approximation". Which should I be using and why?

b. I also saw that you can do a Wilcoxon "test of the mean" when using Analysis-->Distribution-->DeltaScore. I'm assuming that is just to prove that the delta is significantly different from 0. That didn't seem like the correct test to show that interventions improved more than controls. Am I interpreting that correctly?

Thank you all so much in advance for all of your help!!

Brett

Answer 1

If you have a small sample with continuous variables you can calculate the mean difference of both the populations and compare them using a Wilcoxon test.

So the approach would be to compute the difference pre-test and post-test for every subject of the group A and the group B, and we call this variable PrePost.

The you can use the Wilcoxon test to compare the two results you obtained, the PrePost from people in the group A against the PrePost from people of the group B.