Measurement invariance (MI) is a statistical property of measurement which implies that the same construct is measured across different groups, or over time. In high-stakes assessments that measure students’ understanding and mastery of disciplinary concepts, measurement invariance is also regarded as an important facet of test fairness. Statistical methods for evaluating measurement invariance originate from Mantel-Haenszel chi-square statistics and flourish within the item response theory (IRT) framework. In this talk, I will introduce a latent regression with regularization method as an emerging approach for evaluating MI. This new method is particularly tuned for two scenarios: (1) a test measures multiple correlated domains, and (2) a digital-first adaptive assessment with automatic item generation. It will be shown that the new method can simultaneously detect measurement non-invariance in multiple items caused by multiple factors in both scenarios.


Chun Wang is an Associate Professor of measurement and statistics in the College of Education at the University of Washington (UW). She is also an affiliated faculty of the Center for Statistics and the Social Sciences at UW. Her research focuses on psychometric models and methods around educational and psychological measurement, including multidimensional and multilevel item response theory models and applications, cognitive diagnostic modeling and computerized adaptive testing. She is the Associate Editor of the Applied Psychological Measurement, British Journal of Mathematical and Statistical Psychology, and the Journal of Educational and Behavioral Statistics.

https://stat.uw.edu/about-us/people/chun-wang