Application of the dominance and ideal point IRT models to the Extraversion scale from the IPIP Big Five Personality Questionnaire

In the present study, the ideal point and dominance graded response Item Response Theory (IRT) models (respectively: General Graded Unfolding Model, GGUM, Roberts, Donoghue, & Laughlin, 2000; and Samejima’s Graded Response Model, SGR, Samejima, 1969) were applied the Extraversion scale from the Goldberg’s 100 item Big Five personality questionnaire (Goldberg, et al., 2006). Data from 20,000 individuals were used to calibrate the models, and another 20,000 cases were used to evaluate model-data fit and measurement accuracy. Additionally, optimized scales were developed independently for ideal point GGUM and dominance SGR models by removing items that showed poor fit and low discrimination from the original 20-item Extraversion scale. Results revealed that the application of the IRT models did not improve the measurement accuracy of the original Extraversion scale or any of the optimized Extraversion scales developed in this study. The comparison of the model-data fit showed that ideal point GGUM demonstrated a worse fit to the original and optimized Extraversion scales than the dominance SGR model. Moreover, whereas scale optimization significantly improved the data-model fit of the dominance SGR model, attempts to improve the data-model fit of the ideal point GGUM were unsuccessful. Several implications for the IRT models’ application to the existing personality inventories were discussed.