On the questionable soundness of factoring ipsative data: A response to Saville & Willson (1991)

Abstract

Saville & Willson (1991), in this journal, concluded, contrary to many researchers and methodologists, that ipsative scales can be soundly factored and can be viewed as equally valid alternatives to normative scales. We present evidence based on analysis of theoretical and actual data to demonstrate that ipsative scales, although sometimes highly correlated with their normative counterparts, cannot be factored and, when factored, result in predictable factor loading patterns not related to their normative counterparts' factor patterns, and, consequently, will result in invalid factor interpretations. An example is presented where ipsative scores are appropriately and successfully analysed using multinomial statistical techniques. These techniques are more appropriate for ipsative data because ipsative scores contain only categorical information across individuals.