
66 Sample Size Calculations for Clustered and Longitudinal Outcomes
The sample size formulas presented here are simplified versions of Overall
and Doyle [10], which provided equations for calculating z
1−β
to test the
“equal change hypothesis” using simple difference scores or ordinary least
squares (OLS) regression slopes or generalized least squares (GLS) regression
slopes as dependent variables. Let ∆
∗
be the effect size, which is the clinically
or scientifically meaningful difference between treatment means expressed in
standard deviation units. The general power equation for testing “equal change
hypothesis” between two treatment groups is of the form ...