The application of the GSK method to the determination of minimum sample sizes.

Grizzle, Starmer, and Koch (1969, Biometrics 25, 489-503) presented a unified approach for data analysis when the outcome variable is measured on a nominal or ordinal scale. The technique uses a weighted least squares methodology, and hypotheses are tested using asymptotic chi-square statistics. In this paper, we adapt these procedures to the problem of determining the minimum sample size required for an applied research effort, and use the noncentral versions of these chi-square statistics. The results are compared against several procedures widely used in the literature, and are found to concur well with these techniques. As well, some new situations are considered.