TS - Matched Pairs Testing Lesson

Matched Pairs Testing Lesson

Pairing or matching data in some way is indicated in various situations.   In some situations it might be wise to pair siblings or spouses.   An even more common situation arises when we wish to compare experimental results on the same individual before and after treatment.   For example, we might record a student's SAT scores before taking an SAT Prep class with the scores obtained after completing the class.   When data is paired we always rely on the t-test for comparison.  

Pairing is not done simply as a convenience when two samples have the same size, but there must be a "reason" for the pairing, similar to the reasons for blocking in an experiment.   Logically, because the matched pairs are related in some way they are NOT  chosen independently.   Therefore, we will NOT use two-sample inference for a matched pairs design.   We will use a one-sample t test on the DIFFERENCES.

IMAGES CREATED BY GAVS