Research On Item Selection Method Of CAT And CD-CAT With Dynamic Weighted Interval

CAT is the combination of modern measurement theory and computer technology,which application is more and more widely with the in-depth study.Compared with the traditional paper and pencil test,CAT has obvious advantages.In the part of CAT,the item select strategy which is the most intelligence part of CAT is the soul,its quality closely relates to item bank safety questions,test accuracy and so on.Due to the different property and application of CAT,among a number of indicators to measure the quality of the test,some tests focus on measurement accuracy or fairness and others focus on the efficiency,of course there are conflicts between those indicators,but as far as possible consider more indicators after balancing.The research hotspots of cognitive diagnostic computerized adaptive testing is changing from single target tests,which are based on the subject's knowledge state to dual target tests,which are based on both subject's knowledge state and ability.In this paper,new item selection strategies are proposed,which are aiming at the conflict between the usage of item pool and test accuracy,especially the poor usage of item pool of item selection strategies in CAT and CD-CAT,first of all,a new item selection strategy with dynamic weighted interval is proposed for CAT,which can meet the demand of diversity: Firstly,constructing an interval contains the maximum information,there are generally more than one item in this interval,the group of those items are equivalent to a "shadow bank",then set a weight to adjust the size of shadow bank.The interval can balance the item pool usage and ensure test security,the weight can adjust the indicators according to test focus;then the HO-DINA model is used to construct the relationship between the latent traits and attribute mastery patterns of the subjects with CD-CAT:Firstly,using PWKL strategy and dynamic weighted interval strategy to select a group of item,individually and getting the intersection of these two strategies,then dealing with the possible situation of the sets in different ways.The use of intersection achieved dual objectives,that enabled both accuracy of latent traits and attribute mastery patterns,the mixed use of the two strategies is taking the estimation accuracy of subject's knowledge state and ability into account.Monte Carlo simulation shows that both two new types of item selection method's uniformity works prominent while ensuring the accuracy of the test.