Title: Sequential methods for learning under cognitive diagnosis modeling
Authors: Sangbeak Ye - University of Missouri Kansas City (United States) [presenting]
Abstract: Cognitive diagnostic modeling with binary latent attributes classifies each subject into a specific skills profile in a multidimensional binary domain. In the application of e-learning or intelligent tutoring system, the goal is to provide pedagogical resources until the binary latent attribute of the subject consistently corresponds with observational responses that indicate a complete mastery in such a domain under the framework of CDM. The process of transitioning from any state to a complete mastery profile of multiple attributes is viewed as a sequential change-point problem. If each item is assumed to carry different magnitude of stimulus to transition one or more attribute from non-mastery to mastery, irreversibly, the selection criteria of each item may affect the duration until a complete mastery. A variation of item selection methods that adaptively induce the change points and improve the detection accuracy of a complete mastery to gain efficiency was developed. The item selection methods showcase adopting different statistical approaches including Bayesian principles and survival analysis modeling. A simulation study is conducted to compare the performance of the item selection methods.