Based Web-adaptive Testing System

Conventional testing method with paper and pen as instruments is mostly supported by classical testing theory. Its characteristic is that all examinees are tested by the same items. The main defect brought about by conventional testing method lies in the inconsistency between some item difficulty and examinee level, Therefore, the testing result has its inherent high uncertainty. It is incredible to differentiate examinees only by the testing score. Computerized adaptive testing(CAT), which is come forth not long ago and is still developing, is comparatively supported by items response theory. The advantage of CAT lies in that item difficulty is always consistent with examinees'acquaintance with the testing. Testing usually consumes less time, but has more credible result.CAT has been used abroad in America. GRE and TOEFL held by ETS adopt CAT, and CAT is adopted in qualification testing for doctors and nurses. Now the status quo of using CAT in China is as following: the using of CAT is at a small scale; practical CAT, especially designed with B/S pattern is rarely seen. Furthermore, most methods of ability estimation or items-selecting in current CAT are reversely out of date. Hereby there exists much space to study and popularize CAT in our country.On the basis of generally studying the theory related to CAT and technology of database programming and Web design, the author designs and implements a CAT based on Web. Item-selecting method is a key problem in CAT. The research of CAT theory shows that b-stratification is a good item-selecting method. But what is the best way to divide to best suited to the different size of testing and appliment background, which is not recorded? After many researches and experiment comparision, the author separately gives the best number of item sub-database under three different testing sizes. Estimation of examinee ability is also a key problem in CAT. Which method to pick at different phase of testing has not yet fixed pattern, also not recorded in literature. On the basis of fully theory analysis and experiment exploring, aiming at the advantage of different estimation method, this paper separately selects an appropriate estimation method for examinees'ability at different phase of testing.