ABSTRACT

It is observed that a matched study requires a matched analysis, and that much power can be lost if this is not done. Also use of an unmatched analysis for a matched design leads to a bias towards the null hypothesis.