A COMPARISON BETWEEN 1ST-ORDER, 2ND-ORDER, AND 3RD-ORDER FACTOR-ANALYSES IN THE MULTISCALE QUESTIONNAIRE - THE EYSENCK PERSONALITY-INVENTORY

It has recently been shown, using plasmodes, that restricted first-order factor analyses produced very similar results to higher order factor solutions obtained in the usual way. The present study used the responses of 392 subjects on the Eysenck Personality Inventory (EPI) to demonstrate the validity of this assertion with a multi-scale questionnaire. It was found that first-order two and three factor varimax rotations produced almost identical results to third-order factor analyses when comparisons were made using correlations between interpoint distances. This finding was also supported when comparisons were made with other published third-order solutions for the EPI. It is concluded that the greater replicability in structures found by higher order analyses in previous studies are more a function of the inevitable reduction in the number of factors being rotated, than of some mystical power of the factor analytic method itself to reveal such structures only at higher levels. © 1991.