An Empirically Based Method of Q-Matrix Validation for the DINA Model: Development and Applications

Most model fit analyses in cognitive diagnosis assume that a Q matrix is correct after it has been constructed, without verifying its appropriateness. Consequently, any model misfit attributable to the Q matrix cannot be addressed and remedied. To address this concern, this paper proposes an empirically based method of validating a Q matrix used in conjunction with the DINA model. The proposed method can be implemented with other considerations such as substantive information about the items, or expert knowledge about the domain, to produce a more integrative framework of Q-matrix validation. The paper presents the theoretical foundation for the proposed method, develops an algorithm for its practical implementation, and provides real and simulated data applications to examine its viability. Relevant issues regarding the implementation of the method are discussed.