Kernel equating with covariates
(English)Manuscript (preprint) (Other academic)
To equate two forms of a test we need to collect data in such a way that the link between the scales of the two test froms can be esitmated. The traditional approach is to use common examinees and/or common items. In this paper we explore the idea of using variables correlated with the test scores (e.g., school grades, education) as a substitute for common items in a non-equivalent groups design. This is done in the framework of Kernel Equating, and with an extension of the method developed for post-stratification equating (PSE) in the non-equivalent groups with anchor test (NEAT) design. Data from two administrations of the data sufficiency subtest of the Swedish Scholastic Assessment Test (SweSAT), fall 1996 (96B) and spring 1997 (97A), are used to illustrate the use of the method.
kernel equating, covariates, kernel smoothing, equipercentile equating, log-linear models, non-equivalent groups design
Probability Theory and Statistics
Research subject Statistics
IdentifiersURN: urn:nbn:se:umu:diva-32791OAI: oai:DiVA.org:umu-32791DiVA: diva2:305803