Assessing Meaningful Within-Person Variability in Likert-Scale Rated Personality Descriptions: An IRT Tree Approach

摘要

Personality researchers and clinical psychologists have long been interested in within-person variability in a given personality trait. Two critical methodological challenges that stymie current research on within-person variability are separating meaningful within-person variability from (a) true differences in trait level; and (b) careless responding (or person unreliability). To partly avoid these issues, personality researchers commonly only study within-person variability in personality states over time using the standard deviation (SD) across repeated measurements of the same items (typically across days)-a relatively resource-intensive approach. In this article, we detail an approach that allows researchers to measure another type of within-person variability. The described approach utilizes item-response theory (IRT) on the basis of Bockenholt's (2012) three-process model, and extracts a meaningful variability score from Likert-ratings of personality descriptions that is distinct from directional (trait) responding. Two studies (N = 577; N = 120-235) suggest that IRT variability generalizes across traits, has high split-half reliability, is not highly correlated with established indices of IRT person unreliability for directional trait responding, and correlates with within-person SDs from personality inventories and within-person SDs in a diary study with repeated measurements across days 20 months later. The implications and usefulness of IRT variability from personality descriptions as a conceptually clarified, efficient, and feasible assessment of within-person variability in personality ratings are discussed.