Equating Different Test Scores with Landmark Registration Compared to Equipercentile Equating

The overall aim of this study is to propose a new test score equating method based on landmark registration, which has its roots in functional data analysis. A further aim is to compare sum scores and binary optimal scores and examine the proposed method in comparison with equipercentile equating with both simulated data and data from a college admissions test. We used the EG design in both the empirical study and the simulation study. In the simulation study, we examined the behaviour when either difficulty or discrimination differ in the test forms or if there is a large ability difference between the groups taking different test forms. The results indicate that the proposed method worked well. The empirical study suggested that the proposed method could be used in practice. Practical implications of using different test scores in test equating as well as the usefulness of landmark registration was discussed.