Test Theory and Personality Measurement

This article reviews traditional approaches for the psychometric analysis of responses to personality inventories, including classical test theory item analysis, exploratory factor analysis, and item response theory. These methods, which can be called "dominance" models, work well for items assessing moderately positive or negative trait levels, but are unable to describe adequately items representing intermediate (or average) trait levels. This necessitates a shift to an alternative family of psychometric models, known as ideal point models, which stipulate that the likelihood of endorsement increases as respondents' trait levels get closer to an item's location. The article describes an ideal point model for personality measures using single statements as items, reanalyzes data to show how the change of modeling framework improves fit, and discusses the pairwise preference format for use in personality assessment. It also considers two illustrative ideal point models for unidimensional and multidimensional pairwise preferences and shows that, after correcting for unreliability, correlations of personality traits assessed with single statements, unidimensional pairs, and multidimensional pairs are very close to unity.