ON THE ADEQUACY OF SAMPLE-SIZE DETERMINATION FORMULAS FOR 2 INDEPENDENT PROPORTIONS

The performance of six sample size determination formulae for two independent binomial proportions in achieving nominal levels of alpha and beta was evaluated by simulation methods. The formula given in Fleiss (1981) which is not corrected for continuity outperformed other five formulae for beta less-than-or-equal-to 10% for N > 20 and for beta > 10% for N > 100. For N less-than-or-equal-to 20, beta less-than-or-equal-to 10% and for N less-than-or-equal-to 100, beta > 10%, continuity corrected formula of Fleiss, Tytun and Ury (1980) and/or doubly continuity corrected formula of Kramer and Greenhouse (1959) may be the only satisfactory formulae to use.