SOME OBSERVATIONS ON THE SCREE TEST, AND ON COEFFICIENT ALPHA

Abstract

This paper looks into the use of principal components from two perspectives.  Regression lines are fitted to scree plots to see if they might assist with resolving the number of factors question.  Then, the use of the principal eigenvalue and corresponding eigenvector from a classical components analysis is examined, with an eye to identifying best test items. It was found that regression lines may be a useful addition when it comes to determining where an eigenvalue scree breaks. It was also found that the use of item loadings on the first principal component is more likely to identify best items than is the process of looking at item-test correlations.