values and mixed for Monte Carlo-derived critical values. With one Winsorized value per end, for small samples, type I error … most cases, stringently-so for samples of 90 or more) to type I error while maintaining an advantage over Monte Carlo …-derived critical values with respect to type II error. …