Scaling methods for multilevel test data
Several measurement instruments require multiple raters. Examples include an exam in which the responses of each student are rated by two instructors, and the Child Behavior Checklist in which both parents rate the child’s behaviour. Data obtained with these measurement instruments have a multilevel structure, and traditional scaling and item analyses yield inaccurate results with respect to the quality of the measurement instrument. Using flexible measurement models, we develop flexible statistical methods to both construct and assess measurement instruments involving multiple raters. The resulting methods will be made available in user-friendly software.
Prof. L.A. van der Ark & Dr B.J.H. Zijlstra
NWO Research Talent Grant
1 November 2016 – 1 November 2020