Journal article
Bo Zhang, Naidan Tu, Lawrence Angrave, Susu Zhang, Tianjun Sun, Louis Tay, Jian Li
Organizational Research Methods, 2023
Organizational Research Methods, 2023
Semantic Scholar
DOI
Cite
Cite
APA
Click to copy
Zhang, B., Tu, N., Angrave, L., Zhang, S., Sun, T., Tay, L., & Li, J. (2023). The Generalized Thurstonian Unfolding Model (GTUM): Advancing the Modeling of Forced-Choice Data. Organizational Research Methods.
Chicago/Turabian
Click to copy
Zhang, Bo, Naidan Tu, Lawrence Angrave, Susu Zhang, Tianjun Sun, Louis Tay, and Jian Li. “The Generalized Thurstonian Unfolding Model (GTUM): Advancing the Modeling of Forced-Choice Data.” Organizational Research Methods (2023).
MLA
Click to copy
Zhang, Bo, et al. “The Generalized Thurstonian Unfolding Model (GTUM): Advancing the Modeling of Forced-Choice Data.” Organizational Research Methods, 2023.
BibTeX Click to copy
@article{bo2023a,
title = {The Generalized Thurstonian Unfolding Model (GTUM): Advancing the Modeling of Forced-Choice Data},
year = {2023},
journal = {Organizational Research Methods},
author = {Zhang, Bo and Tu, Naidan and Angrave, Lawrence and Zhang, Susu and Sun, Tianjun and Tay, Louis and Li, Jian}
}
Abstract
Forced-choice (FC) measurement has become increasingly popular due to its robustness to various response biases and reduced susceptibility to faking. Although several current Item Response Theory (IRT) models can extract normative person scores from FC responses, each has its limitations. This study proposes the Generalized Thurstonian Unfolding Model (GTUM) as a more flexible IRT model for FC measures to overcome these limitations. The GTUM (1) adheres to the unfolding response process, (2) accommodates FC scales of any block size, and (3) manages both dichotomous and graded responses. Monte Carlo simulation studies consistently demonstrated that the GTUM exhibited good statistical properties under most realistic conditions. Particularly noteworthy findings include (1) the GTUM's ability to handle FC scales with or without intermediate statements, (2) the consistently superior performance of graded responses over dichotomous responses in person score recovery, and (3) the sufficiency of 10 mixed pairs to ensure robust psychometric performance. Two empirical examples, one with 1,033 responses to a static version of the Tailored Adaptative Personality Assessment System and the other with 759 responses to a graded version of the Forced-Choice Five-Factor Markers, demonstrated the feasibility of the GTUM to handle different types of FC scales. To aid in the practical use of the GTUM, we also developed the R package “ fcscoring.”