Detecting latent differential bundle functioning according item features using Saltus model
Yoonsun Jang · Journal of Curriculum and Evaluation · 2025
This study aimed to detect differential bundle functioning by identifying latent classes that respond heterogeneously to items according to item features. For this purpose, the response data of Korean students to the science assessment in PISA 2022 were analyzed using the Saltus model, a type of mixed item response theory model. Item bundles were defined by reflecting item features such as item type, context, competency, and knowledge dimension. According to the results of the study, there was no differential bundle functioning based on item type. In the context dimension, latent groups were found in which both the personal context and regional/national context item bundles functioned advantageously. In the competency dimension, the item bundle related to the scientific interpretation of data and evidence showed differential bundle functioning, and in one latent group, this bundle was found to function disadvantageously. Differential bundle functioning according to the knowledge dimension was identified only in the procedural knowledge item bundle. Moreover, differences in background and educational contextual variables were observed across latent classes in which differential bundle functioning was present. This study aimed to provide empirical evidence for various analysis methods of differential item (or bundle) functioning by exploring latent differential bundle functioning according to item features. Furthermore, the significance of this study lies in proposing an approach that utilizes the results of detecting differential item or bundle functioning not only as evidence for test validity but also as information for instructional support.