The Comparison Between The Results Of Paper-pencil Test And CAT Based On IRT Theory

CAT (Computerized Adaptive Test) is more quickly andefficient than traditional paper-pencil test. Currently, thestudies of CAT mainly focus on three aspects, which containbuilding CAT system, CAT-related algorithm designing andoptimization and combination of CAT and other test theories.However, the accuracy and efficiency of CAT are rarelydiscussed in papers, especially on the features and differenceswhich compared to traditional paper-pencil test under idealsituation.This paper firstly models a real data based on a survey usingN sub-scale of EPQ, in order to study the feasibility of CATunder actual situation and both the similarities and differencebetween CAT and traditional test. Secondly, this paper analysesthe simulated data generated by Monte-Carlo method aim atdiscussing the theoretic and modeling features of both CAT andtraditional paper-pencil test based on IRT (Item ResponseTheory) under ideal circumstances.3conclusions have beenreached as the following.(1) The efficiency enhancement on CAT goes hand in hand witha controlled reduction of the test’s accuracy, the essence ofwhich is a leverage between test’s accuracy and efficiency. Every reduction of the administrated items will cause the dropof test’s accuracy. Organizing the question item is prior toCAT.(2) The additional items will commonly raise the accuracyof both CAT and traditional paper-pencil test. The raise of testaccuracy performs on the lower standard error, which, in CAT,has been set before hand. In CAT, standard error is pre-setparameter and cannot be changed with the enhancement ofaccuracy of the estimated ability value, which draws theconclusion that the standard error cannot present the accuracyof both tests.（3）For people with different ability, the additional itemswill either raise the accuracy or enhance the efficiency of thetest.