DOI: 10.21449/ijate.1929011 ISSN: 2148-7456

Linking high-stakes test forms without common items: The pseudo-equivalent item approach in item pool development

Selma Şenel, Sebahat Gören, Serkan Arıkan, Eren Can Aybek, Tuba Gündüz, Ayşegül Bozdağ Kasap
Show PDF Cite
This study evaluates pseudo-equivalent items as an alternative method for linking high-stakes test forms when common items are unavailable. The findings offer valuable insights for developing sustainable item pools, particularly within Computerized Adaptive Testing (CAT) frameworks. Data were obtained from a national foreign language test administered across six consecutive years. Separate annual IRT calibrations were compared against simultaneous calibration results under a multiple matrix sampling (MMS) design, using fourteen pseudo-equivalent items as anchor indicators. The 2PL model was determined to be the most reliable model for the data. Difficulty parameters from the separate and simultaneous calibrations correlated very highly, with no statistically significant difference between the two sets, and difficulty estimates from Classical Test Theory and IRT were consistently and strongly related, as theoretically expected. Measurement precision was concentrated around average ability levels. The findings provide empirical evidence that pseudo-equivalent items represent a promising approach for linking forms without common items, and that combining this approach with an MMS design provides a feasible strategy for building large, sustainable item pools suited to large-scale and CAT-based testing programs. These results have direct implications for testing organizations seeking to maintain form equivalence and scale stability across administrations without relying on traditional common-item designs, offering a scalable pathway for item pool expansion that preserves measurement comparability over time.

More from our Archive