|
Abstract:
|
Item response theory (IRT) models explain an observed item response as a function of a respondent's latent trait and the item's property and commonly used in behavioral sciences. Local independence, which is a critical assumption for IRT, is often violated during real testing situations, and this violation can severely bias item and person parameter estimates. We propose a new type of model for item response data which does not require the local independence assumption. By adapting a latent space joint modeling approach, our proposed model can estimate relative distances between pairs of items to represent the item dependence structure, which can also be used to identify item clusters in latent spaces. Our approach introduces a new type of item response analysis with opportunities for further applications and extensions.
|
