An Evaluation of Hypothesis Testing Methods for Equating Differences in Kernel Equating
(English)Manuscript (preprint) (Other academic)
In observed-score equating, hypothesis tests of equating differences are helpful in deciding which equating function is suitable. Here, a hypothesis testing procedure for item response theory (IRT) observed-score kernel equating using a Wald test is introduced. Simulations evaluating the Wald test when using IRT and log-linear models are conducted. The test with either IRT or log-linear models is shown to have high power and greatly outperform the Hommel multiple hypothesis testing method. The Wald test is applied to two datasets in both an equivalent groups design and a non-equivalent groups design, showing that the Wald test can provide different conclusions to other hypothesis testing methods in practice.
Probability Theory and Statistics
IdentifiersURN: urn:nbn:se:uu:diva-233486OAI: oai:DiVA.org:uu-233486DiVA: diva2:757290