Title: Score-Based Measurement Invariance Tests for Multistage Testing

Authors: Rudolf Debelak, Dries Debeer

Affiliation: Universität Zürich

Abstract:
Several recent publications addressed score-based parameter invariance tests
for a wide variety of statistical models. In the context of psychometric item
response theory (IRT) models, these tests have been used for the detection of
violations of parameter invariance. This application is particularly
interesting for detecting unfairness in psychological and educational
assessments. A popular application of IRT models in educational assessments
are multistage testing (MSTs) and computerized adaptive tests (CATs). In this
testing methodology, the assessment is adapted to each test taker which allows
more accurate ability measurements in shorter assessments. In this
presentation, we discuss how the ideas behinds score-based measurement
invariance tests can be generalized to MSTs and CATs, with a focus on MSTs.
The proposed statistical tests have been implemented in R and were evaluated
with various simulation studies that provided evidence for the new tests'
power.