Computerized classification testing under the DINA model

The DINA model has been developed to assess whether or not examinees have mastered the latent binary attributes. In recent years, computerized adaptive testing (CAT) under the DINA model has been developed to make cognitive diagnosis assessment more efficient. However its accuracy is not very satisfied. The DINA model and CAT are very different concepts, because the former focuses on classification of latent binary attributes but the latter on estimation of latent continuous traits. In this study, we proposed computerized classification testing (CCT) under the DINA model, because CCT is more in line with the DINA model than CAT. A series of simulations were conducted to reveal the advantages of CCT over CAT. Three independent variables were manipulated: (a) procedure (CCT and CAT); (b) ability distribution of population (normal and uniform). The dependent variables were classification rate and test length required. Item selection was based on the KL information. The results show that CCT yielded a higher correct classification and required a shorter test than CAT, suggesting CCT under the DINA model is feasible.