A0381
Title: Calibrating item response theory models with sparse data
Authors: Shiyu Wang - University of Georgia (United States) [presenting]
Yuan Ke - University of Georgia (United States)
Cong Cheng - University of Georgia (United States)
Abstract: A new statistical methodology framework is presented, tailored to address the intricate challenge of calibrating item response theory (IRT) models under conditions of limited samples and sparse response data. The approach introduces an innovative item parameter estimation method that integrates change point detection techniques, aiming to enhance the robustness and accuracy of IRT model calibration in resource-constrained settings. Situated within the realm of statistical and machine learning methodologies, the proposed approach endeavours to distil valuable insights by unveiling lower-dimensional patterns inherent in the data. To evaluate the effectiveness of our proposed techniques, a series of simulation studies are conducted designed to mimic various characteristics of small-scale assessments. These simulations serve to validate the performance and robustness of our methodology across a range of scenarios commonly encountered in practical applications. Additionally, an in-depth analysis is conducted utilizing real-world data derived from a computer-based classroom assessment, providing empirical evidence of the efficacy and applicability of the approach in real-world educational settings. The outcomes of this research project hold significant promise in advancing the application of IRT in the context of small-scale assessments.
