Abstract
Traditional practice evaluates construct structure through a consideration of some function of the correlations among constructs, for example taking the average correlation (Bannister's intensity) or the size of the first factor. Elements are evaluated in different ways, usually on the basis of interelement distances, as it has been shown (CitationMackay, 1992) that element correlations cannot be relied upon. Perhaps due to these different approaches, element and construct structure are rarely compared. The problem with element correlations is that they change if some constructs are reversed. However CitationCohen (1969) has shown a way of calculating profile correlations that do not change. This approach is shown here as applied to the correlation of elements. Some uses for element correlations are demonstrated.
Notes
*p < .001.