A New Strategy Of Item Selection Of Cognitive Diagnosis Computerized Adaptive Testing

As an important improvement in the test history, Computerized Adaptive Testing (CAT) has unparalleled advantages. With the development of Cognitive Diagnosis (CD) which can provide diagnostic information about students'misconceptions, the Cognitive Diagnosis Computerized Adaptive Testing (CD-CAT) has become an irreversible trend in modern educational assessment. The most commonly item-selection strategies of CD-CAT are Maximum Information method(MI),Kullback-Leibler Information Method(KL) and Shannon's Entropy Method(SHE).Compared in these three item-selection strategies, MI has an advantage on accurate and efficient of the result of estimate Ability of examinees but inefficient at the result of estimating examinees'Knowledge States. On the contrary, the result of SHE is the most accurate and efficient at estimating of Knowledge State but inefficient at estimating of Ability. And as compared with SHE, KL has no advantage on neither Ability nor Knowledge State. Thus, a new method was found for providing more accurate and efficient estimate of examinees'both Ability and Knowledge State at the same time.A new combined strategy was proposed in this study. The combined strategy is that the selected item maximizes the Fisher information at the current value of ability from the candidate items which contain the ten most items minimized the Shannon entropy from the Item Bank. The administered item is optimal for accuracy of Ability estimation and Knowledge State estimation. The new strategy not only maintains a low level of the average of absolute error between the estimated ability and true ability of the examinees but also provides a high accuracy on knowledge states of examinees.