IRT-LR-DIF with estimation of the latent densities
Invited talk by Carol Woods, Department of Psychology, Washington University of St. Louis, USA.
Chair: Wim van der Linden, Tuesday 21st July, 17.15 - 18.00, Uppercroft, School of Pythagoras.
This talk is focused on tests of differential item functioning using multiple-group item response theory . A popular strategy is to compare nested logistic item response models (varying in the constraints) with likelihood ratio tests. Typically, the latent variable is assumed to be normally distributed for both reference and focal groups, but this distributional assumption is not requisite for successful estimation, and may not be tenable for all problems. A more flexible approach will be described, wherein the shape of the latent density for one or both groups is estimated simultaneously with the item parameters as either an empirical histogram or a smooth Ramsay-curve.