Sample size.

Before undertaking a research comparison, investigators may wish to estimate the sample sizes needed to assure that the research is feasible and is worth the effort and expense. Such calculations require several decisions by the researchers: (1) the acceptable level of the type I error (P value), (2) the desired power of the test, (3) the difference between the samples that is considered to be important, and (4) the variability expected among the values to be studied. Some recipes for estimating approximate sample sizes are suggested.