Development and application of graded response model incorporating response times

With the advancements in new technology, tracking examinees’ response processes and gathering such data has become easier. Joint modeling could provide more accurate latent trait estimation as well as more informative feedback. However, most research focuses on dichotomous items, leaving polytomous scoring underexplored despite its prevalence in applied assessments. This study proposes a joint response time, graded response model (GRM-RT) within a hierarchical Bayesian framework, integrating continuous response times with polytomous responses. Through Monte Carlo simulation studies with systematically varied experimental conditions (sample sizes: N=200, 500, 1000; test length: M=10, 20, 40), we evaluated the parameter recovery accuracy of GRM-RT. Results demonstrated robust and stable parameter recovery across all conditions, with improved precision at larger sample sizes and longer tests. Empirical data analysis further confirms the practical application value of the model.