Multidimensional Analysis of Student Growth Using Item Response Theory

• In Event: 35.020 - Excellence in Education Research: Early-Career Scholars and Their Work

Abstract

Many latent traits in the social sciences display a hierarchical structure, such as intelligence, personality, or cognitive ability. Sophisticated IRT models have been recently proposed to model such a hierarchical latent trait structure. However, these currently available models mainly focus on data collected at a single time point, and few systematic efforts have extended these models to measure individual change (across multiple time points). This study focuses on the development, evaluation, and application of longitudinal extensions of the higher-order item response theory (HO-IRT) model using the SEM formulation. Parameter recovery of the longitudinal, HO-IRT model was evaluated via an extensive simulation study. Conditions varied include correlation between latent traits, items loading on each dimension, and number of simulees. For each condition, item, person, and growth parameters are compared when using one of several model formulations or estimation algorithms. Simulation studies demonstrate that little information is lost when separately estimating item and person parameters at each time point as compared to calibrating the computationally and practically complicated complete longitudinal model. A real data analysis shows the feasibility of determining higher-order and domain-specific growth trajectories. We hope this study provides useful statistical tools for reliably reporting and evaluating individual growth on a general (overall) trait and across several, more specific content domains.

Author

• Chun Wang, University of Minnesota