Work

Is pseudo-equivalent groups equating via minimum discriminant information adjustment effective? Evidence from simulation and real data

Year: 2025

Type: article

Abstract: The traditional way to linking different forms of test typically uses common (or randomly equivalent) examinees or items. When common examinees or items are unavailable due to test properties or administrative restrictions, equating with pseudo-equivalent groups (PEG) using background information of the subjects to perform adjustment via minimum discriminant information on the non-equivalent groups has been proposed. To investigate the effectiveness of PEG equating, this study compared the results of PEG equating with those of traditional NEAT equating, using both simulated and real test data. The results indicated that the RMSE and bias generated by PEG equating were slightly larger than those of NEAT equating, while the SEE was smaller under certain simulated conditions. When the correlation between the background information used in PEG equating and the scores to be equated reached approximately 0.75, difference that matters between PEG and NEAT equating did not exceed half of a score unit. This suggests that the equating differences between PEG and NEAT would not have a significant impact, indicating that PEG equating can serve as an alternative to traditional equating designs.

Source: Frontiers in Education

DOI: 10.3389/feduc.2025.1667220

Authors: Xizhen Fan, Minqiang Zhang, Hengli Peng

Institutions: Changsha Normal University, South China Normal University, Beijing Language and Culture University

Language: English

Topic: Psychometric Methodologies and Testing

Subfield: Management Science and Operations Research

Field: Decision Sciences

Domain: Social Sciences

SDG: Quality education

Open Access status: gold

Estimated APC paid: $2,353