A Comparison of the Orthogonal and the Oblique Factor Structures of Correlation Matrixes of Individual Items and of Composites of Items (Subtests) Derived from a Standardized Affective Measure

Abstract

For a 90-item standardized affective measure containing six factor scales (with 15 items per scale) that in its original form of 150 items (25 items per subscale) had undergone three item-analysis studies (with two representing cross-validation efforts) to achieve a relatively higher degree of homogeneity than that in the orginal scales, the major purpose of this investigation was to compare both an orthogonally and obliquely rotated factor structure of a correlation matrix of individual items with that of sets of composites of five items (three subtests per factor scale). The two groups of tenth and eleventh grade students employed in this study came primarily from middle-class families: (a) a suburban Anglo sample of 337 subjects (170 males and 167 females) and (b) a rural Anglo sample of 146 (74 males and 72 females) with fewer than two percent of the participants in either sample representing minorities. The orthogonally rotated factor solution was achieved from use of the varimax procedure, and the obliquely rotated factor solution was effected from employment of the promax method. The two major conclusions suggested by the data analyses were that (a) both an orthogonally rotated (varimax) and an obliquely rotated (promax) factor solution can be expected to yield meaningful psychological dimensions that can be interpreted as corresponding to the scales of the instrument irrespective of whether an item correlation matrix or subtest correlation matrix has been analyzed and (b) the orthogonally rotated and obliquely rotated factor solutions afford comparable structures that permit inferences regarding what are readily interpretable or identifiable latent traits underlying a test that has been designed to operationalize a given psychological domain. The outcomes of these exploratory factor analyses are being used in several confirmatory factor analyses with various data bases to test a number of alternative hypotheses regarding the factorial structure of the affective measure employed.