2022 Early Career Award
In standardised educational testing, items are repeatedly used. Some items may get leaked after exposure in a few test administrations, and some test takers obtain access to the leaked items and gain an advantage in future tests. In this talk, we propose statistical models and methods for detecting item preknowledge in educational tests. We consider two different settings: (1) The detection of leaked items and test-takers with preknowledge based on item responses and response times from a single test, and (2) the online detection of leaked items based on sequentially collected data. We view the first problem as a two-way outlier detection problem for multivariate data and propose a latent variable model and associated compound decision theory to detect the two-way outliers. We view the second problem as a multi-stream sequential change detection problem and propose a compound decision theory to detect changed streams quickly. The proposed methods show superior performance under real and simulated settings.
About the Speaker
Dr Yunxiao Chen is an assistant professor at the London School of Economics and Political Science. His research focuses on the development of statistical and computational methods for solving problems in social and behavioural sciences, under three interrelated topics, including (1) large-scale item response data analysis, (2) measurement and predictive modelling based on dynamic behavioural data and (3) sequential design of dynamic systems with applications to educational assessment and learning. He has published in leading journals in psychometrics, statistics and machine learning, including Psychometrika, British Journal of Mathematical and Statistical Psychology, Journal of Educational and Behavioral Statistics, Journal of American Statistical Association, Biometrika, and Journal of Machine Learning Research.
Before joining LSE, Dr Chen was an assistant professor in the Department of Psychology and the Institute for Quantitative Theory and Methods at Emory University. He completed his PhD in Statistics at Columbia University in 2016.