2025 Volume 73 Issue 1 Pages 14-27
In recent years, many municipalities have administered assessment programs in elementary schools in order to identify longitudinal trends in children's learning abilities. Most of these programs have used multiple test forms for each grade; no common items were used. In order to construct a common vertical scale, another test, an "anchor test", has to be administered to monitor groups that have small sample size. In Study 1, an anchor test for mathematics was constructed from existing test forms. The existing data and the anchor test data were then combined and analyzed using item response theory (IRT). Scales from the test forms for grades 2 to 6 in five elementary schools were aligned by two different methods: concurrent and separate calibration. A comparison of these two methods suggested that concurrent calibration could be an alternative method when separate calibration results are unstable due to small sample size on the anchor test. In Study 2, test data from different academic years between grades 2 and 6 were organized according to student. Item response theory analysis and the scale-aligning process using concurrent calibration revealed latent ability estimates for students by academic year. Furthermore, longitudinal trends in ability could be classified into distinctive patterns.
