Key points are not available for this paper at this time.
Abstract A test for change in composite reliability as a result of scale revision or modification is proposed. The approach can be used either in a one-tailed manner to test for increase (or decrease) in reliability or in a two-tailed fashion to test for a difference in the composite reliability coefficients as a consequence of dropping or adding one or more components. The procedure is useful for purposes of examining whether a revision changes the population reliability of a scale undergoing development. The method can also be utilized for addressing questions about shortening of long scales. The outlined approach is illustrated with a pair of examples.
Raykov et al. (Tue,) studied this question.