Statistics Department Seminar Series: Zhiliang Ying, Professor of Statistics, Columbia University

Abstract:
Measurement theory has played a foundational role in educational and psychological assessment. Traditional methods are primarily based on latent variable models, which could be confounded with dimension reduction and goodness of fit. Furthermore, these methods are not applicable to modern complex data such as process data obtained from PSTRE (Problem solving in technology-rich environments) items. This talk will first introduce a new method for addressing the first issue, i.e. sufficient dimension reduction with adequate goodness of fit. It is followed by a second new method for handling process data from PSTRE items. These new developments will be applied to data in educational assessment and psychological evaluation.