Item-Analysis Procedures for Enhancing Validity of Existing Personality Scales

Abstract

Item-analysis procedures were used to reduce a 51-item hostility scale to a 20-item version with equivalent reliability and convergent validity, yet freer of desirability variance. Evidence presented in a multitrait-multimethod matrix (8 traits by self-ratings of trait adjectives and by item endorsements) indicated an enhancement of discriminant validity in the shorter form. The technique is presented as a means for improving the validity of scales developed by other earlier scale construction methods.