Effects of item type and estimation method on the accuracy of estimated personality trait scores: Polytomous item response theory models versus summated scoring

Abstract

Despite an increased use in item response theory (IRT)-based personality testing there is little research documenting whether trait estimations are actually improved over those derived via simply summated scoring according to classical test theory (CTT). In this study personality scale validity was compared using a variety of estimation methods (CTT, adjusted-CTT, SGR, GGUM) and item types (monotonic vs. non-monotonic) for the traits of conscientiousness and extraversion. Regardless of item type or estimation method, trait estimates were highly correlated. Using job performance ratings as an external criterion within the nomological network of these traits, model fit was not related to scale validity, and all estimation procedures resulted in comparable validity coefficients. Implications are discussed.