Conscientiousness at the workplace: Applying mixture IRT to investigate scalability and predictive validity

Mixture item response theory (IRT) models have been used to assess multidimensionality of the construct being measured and to detect different response styles for different groups. In this study a mixture version of the graded response model was applied to investigate scalability and predictive validity for a Conscientiousness scale in a career development context (N = 9283). A four-class solution yielded the most interpretable results. The classes differed mainly with respect to their scores on the subscales Perfectionism and Concentration. Results showed that Conscientiousness may be qualitatively different for different groups of persons and that the predictive validity of the test scores improved for persons in different classes as compared to fitting a unidimensional IRT model. Implications of this study for personality assessment are discussed.