The Item Factor Structure of the Personality Research Form

Abstract

Multiscale personality inventories have rarely, if ever, demonstrated factor structure broadly consis tent at the item level with scale keys. An item fac tor analysis of the 352 items of PRF-E was under taken to evaluate the extent to which PRF items de fine separate and distinct factors corresponding to keyed scales. A matrix based on 352 items and 214 subjects drawn from 31 colleges was decomposed according to the Eckart-Young theorem, the factor loading matrix was formed, and then was rotated to an orthogonal target matrix of the 22 PRF-E scales. Inspection of the rotated matrix showed that only 2 of the 352 items failed to load in the keyed direc tion. The mean loading of items on their scale fac tor was .38; the mean loading of non-scale items was .09. This strong tendency for scale items to load more highly than non-scale items was also re flected in the majority of scale items being among the 16 highest loadings on each scale factor. For three scales, Dominance, Harmavoidance, and Or der, the scale items were the 16 highest loaded items. The observed item factor structure of the PRF is interpreted as being attributable to the methods of construct definition and item selection used in the construction of the PRF.