Empirical Prediction Models for Training-Group Assignment
本研究利用双交叉验证程序，基于预测训练时间建立多元回归模型，将学生分配到两种训练条件，结果显示最优分配比随机分配节省47%训练时间，比错配分配节省53%。
A double cross-validation procedure consisting of 58 subjects in Sample 1 and 40 subjects in Sample 2 was used to develop combined sample empirical prediction models for assigning students to one of two training conditions based upon predicted training time for each type of training. The effectiveness of these multiple regression equations for training-group assignment was tested in a subsequent study. In the study evaluating the assignment procedures, 40 students were assigned to either a fixed-difficulty or adaptive training situation based upon their shorter predicted training time (optimal assignment). An additional 40 students were assigned to a training group based upon their longer predicted training time (mismatched assignment). Finally, a control group of 40 students was randomly assigned to one of the two training conditions. Using multiple linear regression models and predicted training times to match students to training alternatives resulted in a 47% savings in training time compared to random assignment and a 53% savings compared to mismatched assignment. Variability in training time was reduced approximately 40% using the empirical prediction models for assignment.