Abstract
The two stage sample size determination method by C. Stein has been widely utilized to determine sample size. The normality of the parent population is assumed in Stein's method. However, the normality assumption does not always hold good for the practical application. It is known that Stein's method gives unreasonable result for the non-normal population. In this paper, it is discussed how the sample size derived from Stein's method is affected by non-normal population represented by the first four terms of an Edgeworth series, and a table of correction coefficients for Stein's sample size is given. A simple method of the non-normality correction for the sample size is also given by dint of Fisher-Cornish approximation.
