THE EFFECT OF RETAINING PROBABILITY VARIATION ON SAMPLE-SIZE CALCULATIONS FOR NORMAL VARIATES

Skalski (1992, Biometrics 48, 877-882) assumed a constant retaining pr obability p for all experimental units and found that the traditional ad hoc procedure of escalating sample size by the multiplicative facto r lip tends to underestimate the required sample size when p is small. Because this retaining probability p often depends on characteristics of different experimental units, however, p may vary from one unit to another. In this report, we have allowed p to vary among different ex perimental units. We have noted that when the retaining probability p has a small mean but a large variance, the underestimation of the requ ired sample size for a desired precision by use of the traditional adj usted sample size formula can be even more serious than that demonstra ted by Skalski.