In the CAT context, item selection is one of the main issues, as well as the stoping rule which investigated in the paper presented last week. Moreover, this paper also in the context of cognitive diagnostic assessment. First, this paper introduce the fusion model (full model) which is the more complex model in the CDM family. Then, the authors clarified the assessing method of item information in CD-CAT context (purely KL information and PWKL information). Based on the PWKL information and take into account the exposure control, this paper proposed two item selection methods which not purely on the item information but also considered the exposure control, which is a couple of variables with trade-off relationship. These two model with a random part and so the exposure also be modeled. Two simulation study with the same item parameters and person parameters were conducted to show the propretice of the models and compare them to other commonly used models. The results show that the proposed model can take a balance in psychometric precision and item exposures. Future research directions also be discussed.
Due to the unfamiliar with the CDM, there are some issues call for further reading in the future for myself. In addition, though the fusion model selection with the statement from the author that there are two empirical study available up to now, how about the other models? Will the proposed method show the same or equivalent propertice in them?
I don't know whether the content balance issue in CDM is still need to take into account. But in a intuitive level, is it need to be take care when we try to control the item exposure? This paper did not show the coverage of attributes of these two item selection methods. Although the recovery rate of the attributes is very good, how about the situations involving more attributes but not all of them are with amount of items?
Depart from the future directions which raised by the authors, a immediately available research questions might be involving the stop rule which proposed in the paper last week with the proposed item selection methods.