Alternative methods for the factor analytic treatment of ipsative scores, with particular reference to the Edwards personal preference schedule

Abstract

This study investigated a method of removing the linear dependency and forced negative intercorrelation of ipsative scales that make them mathematically unsuitable for inclusion in a factor analysis with other variables. One scale of the Edwards Personal Preference Schedule was deleted and an inverse transformation of the remaining scales was computed. Three intercorrelation matrices were then formed using different combinations of the original EPPS scales, the inverse EPPS scales, and a basic set of 49 other personality measures. The three intercorrelation matrices were factored and rotated toward a simple structure. The results of these three factor analyses were examined and reported separately and then compared in terms of factor definition, factor clarity, and factor structure. The major finding was that using the inverse did not substantially improve the factor solution over the method of total exclusion, but was superior to using the ipsative scales in their original form.