An application of the empirical Bayes estimation to improve the reliability of subscores

In this paper, we show a subscore augmentation procedure that improves the reliability of subscales. The approach uses empirical Bayes estimations. This a generalization of Kelley's formula. The estimates are based on the information from other related scales on the test. We describe the procedure and we apply it to a multidimensional test. The reliability of the subscores increased, and the improvement could be considered equivalent to a 58.06% increase in the length of the test.