Wilcoxon or t-test question
Message posted by Donna Blair-Richardson (via 204.108.241.14) on March 4, 2002 at 4:38 PM (ET)

I have two groups of students-one large (120) and one small (20) for which I matched on several variables such as SES, IQ, gender, teachers. When I finished matching, I only had nine students I could match on all these factors. I wanted to do a paired t-test, but thought I might need to do a Wilcoxon because I have a small sample and cannot be sure it follows a normal distribution. However, three of the nine matching scores are the same, so there is a difference of 0. Also, the scores are ranks themselves such as 0, 1, 2, 3, and 4. Should I even try to perform a statistical test? I'm not sure whether I have enough data to tell me anything worthwhile even if I were to conduct the test. Thank you for your help.