نوع مقاله : مقاله پژوهشی
نویسندگان
1 دانشیار گروه مطالعات برنامهریزی درسی و سنجش و اندازهگیری، دانشگاه خوارزمی، تهران، ایران
2 دانشجوی دکتری رشتۀ سنجش و اندازهگیری، دانشگاه علامه طباطبایی، تهران، ایران
چکیده
کلیدواژهها
عنوان مقاله [English]
نویسندگان [English]
This study aimed to develop and evaluate an item response theory (IRT)-based framework for equating high school final examination scores used in national university admissions, to ensure fairness and control differences in test difficulty across years. A descriptive–psychometric design based on Monte Carlo simulation was used. As the study was simulation-based, data were generated under controlled conditions rather than drawn from samples. One hundred replications were performed for each scenario within a nonequivalent groups design with common items, across three sample sizes (20,000, 100,000, and 500,000 examinees) and four test lengths (15, 20, 25, and 30 items). Anchor items were assumed to represent the overall test construct and content, with parameters generated from the same distribution and implemented in a mixed-format design reflecting national exams. Test reliability was examined using test information, showing highest measurement precision in the ability range typical of most examinees, confirming adequate psychometric adequacy of the model. The performance of several IRT-based and classical equating methods was evaluated. Methods based on the generalized partial credit model—especially concurrent calibration and the Stocking–Lord procedure—showed the lowest equating errors (RMSD). In contrast, the mean–mean, mean–sigma, and robust mean–sigma approaches yielded larger errors, particularly with shorter tests or smaller samples. Increasing the proportion of anchor items improved equating accuracy. Overall, IRT-based equating promotes more equitable and interpretable score comparisons in university admissions.
کلیدواژهها [English]
)