Sample size determination for comparing two therapies taking both the α (type I or false positive) error and the β (type II or false negative) error into consideration

摘要

The sample size required for determination of a given degree of difference between two therapies or modalities A and B being compared with each other, supposing that the sample sizes are equal, is given by Here Cα/2 is the upper α/2 percentile point of the standard normal distribution, C1−β is the abscissa cutting off the proportion 1−β in its upper tail and β in its lower tail; pA and pB are the probabilities of favorable response to A and B; if p̂A and p̂B are the corresponding measured proportions, then p = 12(p̂A + p̂B); and the Q's are 1 minus the p's. When it is desirable to use unequal size groups, the sample sizes n1