Application of IRT equating on high-stakes testing in Applied Universities of Teacher Education. Errors and error-analysis in the consistency and stability of pass/fail decision in tests with different sample sizes
In psychological testing, or exams, three measurement aspects are important: tests have to be valid, reliable and fair for the test takers. Meaning that the chance of passing on an exam only needs to be related to measured ability of a test taker. To have a fair cut-score on each exam, tests need to be empirically equated, preferable with Item response theory (IRT), and the cut-score transferred from one exam to another. Unfortunately, IRT has large sample requirements. Besides sample size, also other aspects influence the estimation precision in statistical equating and in transferring the cut-score, such as, test length, samples ability, anchor test etc. This research is about studying these effects.
The main research question is ‘Which of these sources of errors affect most the final cut-score, which has been determined by means of statistical equivalence, and is it possible to minimize those sources of error in such a way that the final cut-score that is transferred, has a small margin of error?
Prof. T.J.H.M. Eggen & Dr N.D. Verhelst
Vereniging Hogescholen, project 10voordeleraar
1 November 2016 – 1 October 2020