Outcome in Two Large Sample Studies of Factorial Similarity under Six Methods of Comparison

Abstract

In studies of test bias evaluating the cross-group similarity of factor analytic results, a variety of methods of measuring factorial similarity have been employed. Six of these methods were compared with regard to outcome based on two large data sets, one for an intelligence test and the other for a personality test. All indexes yielded comparable results regardless of whether factors had been derived from subtests of the intelligence test or items of the personality scale. Conclusions would have remained constant in each study regardless of the index employed. When used to compare factors determined at random, all indexes yielded comparable results leading to a conclusion of dissimilarity, though the values of the salient variable similarity index were uncomfortably high.