Studies Of Dual-objective Computerized Adaptive Test With Single-calibration Process

In order to simultaneously obtain the general ability scores as well as detailed knowledge state information used to improve learning and teaching and so on,the dual-objective cognitive diagnostic computerized adaptive testing(Dual-objective CD-CAT)has been proposed,which simultaneously estimates the information on both the general ability and knowledge state under the framework of traditional CAT.According to the calibration process of item parameters,the dual-objective CD-CAT can be divided into dual-objective CD-CAT adopting dual-calibration process and dual-objective CD-CAT adopting single-calibration process(ie Higher-order CD-CAT).Different from the former calibrating the same item through the two models(IRT model and CDMs model)which the underlying latent construct is dramatically different,the latter just calibrates a set of parameters for each item.Therefore,the Higher-order CD-CAT is more reasonable,and its cost in the construction of item bank is lower than the dual-objective CD-CAT which adopts dual-calibration process.However,the item selection method which is applicable to the Higher-order CD-CAT has not yet been proposed,the item selection of this CAT system depends on the traditional item selection methods used in the single-objective CD-CAT.These methods only considered the knowledge state and hence ignored the estimation precision of the general ability,which would violate the purpose of the dual-objective CD-CAT.Therefore,this research was designed to propose a new item selection method for the Higher-order CD-CAT,which simultaneously consider the general ability and knowledge state,the main idea is as followed: Under the framework of the higher order CDMs,the knowledge states of examinee was governed by the general ability,and the responses were modeled by the knowledge states,hence it is reasonable to hold that the items responses were modeled by a combination of the general ability and knowledge states.The present article incorporated the posterior probability of the combination into the Shannon entropy method and finally developed the HO-SHE method,which simultaneously consider the general ability and knowledge state and is applicable to the Higher-order CD-CAT.On the other hand,although the polytomously-scored items were widely used in practice,the present dual-objective CD-CAT could not analyze these items,which seriously limited the practical application of the dual-objective CD-CAT.In light of this,another purpose of this article is to propose the polytomously-scored Higher-order CD-CAT to analyze the polytomously-scored items based on dichotomously-scored Higher-order CD-CAT.The results of Monte Carlo simulation are as followed: Under the framework of dichotomously-scored Higher-order CD-CAT,the HO-SHE method achieved more accurate estimation on both the general ability and knowledge states across the different conditions compared with the traditional methods,especially under the HO-DINA model;Meantime,the item exposure rate of the HO-SHE method was smaller than the traditional methods,and showed the more uniform utilization of the item pools.Under the framework of the polytomously-scored Higher-order CD-CAT,all of the item selection methods produced equal or smaller MSE value than 0.215 and higher PCCR value than 0.985,which indicated that the polytomously-scored Higher-order CD-CAT work well on the polytomously-scored items;In addition,analogous to the results of the dichotomously-scored items,the HO-SHE method produced higher estimation precision of person parameters than traditional methods,which demonstrate that the new method can be used in different Higher-order CD-CAT systems.In conclusion,the present article provided a new and more effective method to achieve the true purpose of the dual-objective CD-CAT.