Variable-length computerized adaptive testing for cognitive diagnosis

Cognitive diagnosis has become increasingly popular in educational assessment, because it gives diagnostic information about students’ learning status. In contrast to a summative score that is provided by traditional item response theory models, a profile of individual attributes is obtained from cognitive diagnostic models. In recent years, there has been an increasing interest in developing computerized adaptive testing (CAT) for cognitive diagnostic models and most of the algorithms use a fixed length termination rule. This study aims to develop a variable length CAT for cognitive diagnosis and evaluate its performance using simulations. The simulation results showed that the two proposed termination rules worked very well and the variable length CAT was feasible.