(xI

C. Example In a hypertension trial comparing a test treatment versus a control with respect to blood pressure change from baseline, we require d = 4 mmHg and a = 0.05. From previous information, (J is guessed as about 10 mmHg, although this guess may be in error. The first step in applying the foregoing method is to select a value of nl, the size of the initial sample. For this, it is helpful to know how large the final sample per group, N, would have to be if the assumed value of (J happened to be correct. This value is

N = 2 [ : tJ2 = 2 [ ~ X 2J2 = 50 where ( = 2 is taken from the normal approximation to the ( distribution. Suppose that nl is taken as 30. With 30 observations per group, we usually can estimate (J reasonably well.