The Maximum Priority Index Method with Flexible Content Balance Constraints in Computerized Adaptive Testing

摘要

The maximum priority index (MPI) method is a new heuristic approach. This method is used for severely constrained item selection in computerized adaptive testing. It is able to accommodate various nonstatistical constraints simultaneously, such as content balancing, exposure control, answer key balancing, and so on. It allows for trade off among constraints. The MPI method can be considered as a variant of the maximum information method. The MPI method in its current form is limited to constraints that are in the form of upper bounds. For the flexible content balancing constraint involving a lower bound and an upper bound, the MPI method needs to be modified. Based on the two-phase item selection framework, in this paper we use a new function fk=(vk-xk)/vk to simplify and modify the MPI method. Compared with the weighted deviation modeling method, it leads to this method is feasible.