Cronbach's alpha is industry standard for assessing reliability. It measures consistency of items forming a scale, which is useful but could be misleading. The problem is that the items could all measure a composite of several latent traits; hence the scale may not be very useful as an indicator of a construct in spite of a high value fo alpha. A more appropriate measure need to reflect true homogeneity of the items in a scale, simply speaking if they measure a common factor. This is probably going to replace alpha in test construction and assessment of tests, but the process is fairly slow, and the practical implications so far not well known. How many test scales need to be "purified" and will they then be more useful in practical applications?
Reference
Schweizer, K. (2011). On the changing role of Cronbach's alpha in the evaluation of the quality of a measure. European Journal of Psychological Assessment, 27, 143-144.
No comments:
Post a Comment