Statistical Study Of Adaptive Topic Selection Strategy Based On 3PLM

Computer adaptive test is a new type of test which is based on the item re-sponse theory to establish the question bank,choose items automatically by the computer according to the ability of the subjects,and estimate the ability of sub-jects at last.Computer adaptive test has been widely used in foreign countries,such as TOFEL,GRE,and other large examinations.The most important part of the computer adaptive test is the strategy of topics' selection.In recent years,many scholars have made a deep research on the strategy of topics' selection,because it is related to the accuracy,safety,efficiency and reliability of the test.This paper introduces the basic principles and the process of computer adaptive test,and intro-duces some commonly used strategies and their advantages and disadvantages,such as the strategy of maximum information,the strategy of capacity matching and the strategy stratified by discrimination.Then we constructed an improving strategy of topics' selection which consideres both exposure rate and difficulty.Next,we simulated a question bank by Monte Carlo method and estimated the ability of sub-jects.Finally,this paper simulates the comparison of various selection's strategies by Matlab and evaluates their strengths and weaknesses.