Factor Analysis of the Personal Profile System

Principal components extraction with orthogonal and oblique rotations tested construct validity for the Personal Profile System. MOST-LIKE endorsements of 96 behavioral descriptors were coded with 4, LEAST-LIKE with 1, and unendorsed with 2.5. Descriptor data from 1,045 senior noncommissioned Air Force officers were normalized. Four factors accounted for 85% of total variance, with 19 descriptors loading significantly on two factors and the remaining 77 on just one factor. The measure of sampling adequacy for every descriptor exceeded .94. One factor for the varimax-rotated (best) analysis was bi-scalar, loading on Steadiness and Compliance descriptors; a second resembled Influencing, a third loaded almost exclusively on Dominance, and a fourth did not contain a nonchance number of loadings for any single theoretical dimension. All descriptors loaded on at least one factor at .30 or higher, accommodating an acceptable theoretical degree of psychometric and measurement properties and indicating four-factor relevance. Results do not completely justify previous Personal Profile System publisher claims.