The Study Of Using HCI To Detect Construct Error In Cognitive Attribute Hierarchy

Many cognitive diagnostic models are developed based on Q matrix, Q matrix is uniquely decided by pre-constructed cognitive attribute hierarchy. For these cognitive diagnostic models to yield inferences about students, the attribute hierarchy in the domain of interest must be specified correctly because it is the basis of explaining models outcome. Cognitive attribute hierarchy is constructed by experts and teachers after they analyze theoretically to attributes and their interrelationship in the domain of interest. Three questions must be thought when researchers are constructing cognitive attribute hierarchy. The first is whether each attribute itself is scientific, the second is whether overall attributes is integrated, and the last is whether the interrelationship between the attributes is correct. The science of attribute itself must be analyzed by experts in the domain of interest. After experts construct cognitive attribute hierarchy based on theory, the hierarchy consistency index can be used as assistant method to detect integrality of overall attributes and correctness of their interrelationship. By comparing an observed examinee response vector to the Qr matrix attained from attribute hierarchy, the HCI can be used to assess whether the cognitive attribute hierarchy represent really examinee's cognitive process when solving test items. That is, HCI compute each subject's HCI respectively to demonstrate whether each examinee uses different cognitive rule from what the Qr matrix indicate when solving test items. The mean and standard deviation of the HCI can be used as indicators of the overall model-data fit. A high mean and low standard deviation suggest that the item response vectors fit the model well. This study use overall subject's HCI mean and standard deviation to examine integrality of overall attributes and correctness of their interrelationship. By analyzing different error interrelationship structure between attributes and attribute missing structure and comparing HCI sensitivity in all kinds of error structure, I give some operational proposal for teachers how they can examine constructed cognitive attribute hierarchy base on theory in educational cognitive diagnosis test and teaching practice.