Introduction

Goal: How can we determine if a treatment changes an outcome for each unit?

Background

Take the following question: Does a topical treatment lower a patient’s rating of pain? Suppose we have before and after ratings for each user. Since each user may have a different level of pain going in, taking the overall difference of means may not show any significant change. Instead, we can look at differences per-patient, and see what that looks like.

As before, the parametric solution (paired t-test) makes a few strong assumptions about the data, mainly the assumption of normality. This chapter describes alternative tests we can use instead.