Latent variable and latent class models for binary item response data in psychometric tests (4pm)

Irit Aitkin

Department of Mathematics and Statistics, University of Melbourne

This talk overviews the classical 2PL model for binary item  response data in large scale psychometric testing, and discusses extensions of the model to multilevel designs and to mixture distributions for the latent variable. Serious problems arise with missing data at the upper level, and with model comparisons using the likelihood ratio test. These need a fully Bayesian treatment for adequate analysis.

Dr. Irit Aitkin is a Senior Research Fellow in the  Department of Mathematics and Statistics at the University of Melbourne. She trained in Statistics at Tel Aviv University, then worked as a lecturer and in statistical consulting at the University and for Tel Aviv Medical Center. She has since had appointments at NCEPH in Canberra, in the School of Public Health at Curtin University of Technology, the Statistics Department and Medical School at the University of Newcastle (UK), the Department of Mathematics at the University of Western Australia, the Department of Applied and Engineering Statistics at George Mason University and now works in Melbourne on a grant from the US Department of Education and also with her husband Murray on an ARC grant. Dr. Aitkin’s research interests include identifiability and inference in item-response models, random effects and latent class modelling, and Bayesian inference.