Effect of Personality Item Writing on Psychometric Properties of Ideal-Point and Likert Scales

摘要

The present study was designed to investigate personality item-writing practices and their effect on the psychometric properties of personality items and scales. Personality items were developed based on ideal-point and dominance models, analyzed using the generalized graded unfolding model, and empirically classified as having an ideal-point or dominance form. Results suggested that writing dominance items were slightly easier (more successful) than writing ideal-point items, but this varied slightly by personality dimensions. Of 3 ideal-point item writing tactics, the "neutral" tactic was least successful; success writing "double-barreled" and "average" ideal-point items was comparable to that of dominance items. Three personality scales were then constructed using successful ideal-point and dominance items. Scales constructed using ideal-point items had substantially inferior psychometric properties, including lower score reliability, lower correlations with important criteria, and mixed test information results. However, lower predictive validity of ideal-point scale scores may be due to lower reliability of the scores. Practical and methodological implications were also discussed.