Abstract
This study investigated an empirical method for setting optimal cutting scores for a criterion-referenced archery test. The classification-outcome probabilities and approaches to validity suggested by Berk were utilized. Pretest scores were obtained on 35 uninstructed college-age women on six ends (six arrows each) from 20 yards (18.3 m) after an unrecorded warm-up end. Posttest scores were after 15 weeks of instruction. Score distributions were the primary determinant for accurately classifying students as true mastery and true nonmastery. Accuracy is a function of the amount of overlap between distributions. Using the point at which the distributions overlapped, classification accuracy was estimated. Probabilities associated with 80 points were p(TM) + p(TN) = .83 and p(FM) + p(FN) = .14. Scores above and below 80 points had lower probabilities of classification accuracy. Reliability estimated using Kappa was .59. Statistical validity of the cutting score (phi) was .68.