Predictive performance of psychological tests: is it better to use items than subscales?

Machine LearningPsychometricsCross-validation