2014 Volume 10 Issue 1 Pages 69-80
We propose a standard error of measurement for linked scores and the index of reliability for linked scores in order to enrich the indices used to analyze the feasibility of linking. Our proposed indices can be easily estimated by using basic statistics such as the variance of test scores, coefficient of reliability, and correlation coefficient. When a data collection design is a single group design, the proposed indices can be readily estimated by basic analysis of test results. Even with an equivalent groups design, if the inter-test correlation can be predicted then the proposed indices can be used. Furthermore, we present numerical examples that examine the feasibility of linking from a reliability perspective by using the proposed indices together with the results of actual large-scale tests. Through the examples, we showed that the proposed indices are useful for evaluating the reliability of linked scores and that they provide various indications for making decisions about the feasibility of linking.
