This event is being organized by the NCME Contemporary Issues in Scaling, Linking & Equating SIGIMIE.
Test equating allows us to identify scores on different tests (typically) taken by different groups of students that can be treated as equivalent. Common items that are included in both tests may be used to facilitate this task.
Many different equating techniques have been described in the psychometric literature. However, there is no clarity on which methods work best, under which circumstances.
To tackle this problem, we carried out the largest ever empirical study on this topic. We used data from 1,000 real tests to compare over fifty different equating methods. For each of the 1,000 tests in the analysis, a pseudo-tests approach was used to mimic real equating scenarios. Each test was broken into two overlapping test forms and the students were (non-randomly) assigned to each one. We varied the anchor test length, the number of students “taking” the pseudo-tests, and the ability difference between the groups. Then, we evaluated the equating techniques by comparing their resulting equating functions to the ‘true’ relationship.
Our results suggest that just six equating methods should provide good equating performance under different combinations of group ability differences and sample sizes.
Presenters:
- Tom Benton & Matthew Carroll, Cambridge University Press & Assessment
Please note that the views and opinions expressed during the presentation are solely those of the presenters and do not necessarily reflect the views of the NCME SLE SIGIMIE and/or its committee members. Our goal is to foster open discussions and critical thinking within our field.