Designing and Implementing of the Calibrated Item Bank for Nationwide Entrance Examination

This article reports the process of establishing and developing calibrated item bank based on item- response theory. 30 differential calculi multiple choice items were experimentally administered on a sample group of 3409 junior high school students. Maximum Test Information Function over theta scale on point 0.95 was 43.21. On this point of the theta scale, the ability was estimated with minimum measurement error. After specifying test item's parameters and calibrating the test items into a pool, Fast-test Professional Software, a linier sub test was developed with items discrimination of 2 £ ai and local parameters1 2 i £ b £, respectively. At last 5 items were selected among the 30 bank's items for the sub test and maximum subtest information function over the theta scale at point 1.35 was 12.48, and hence, information function level decreased 3.46 times because of the subtest shrinkage from 30 to 5 Items. Difficulty level of the subtest increased 0.4 standard deviation units. Furthermore, in order to run computerized adaptive testing, one post simulation study was lunched. This simulation study had shown at most 8 to 10 items were needed to measure individual's ability with error probability less than 0.01 comparing with the origin test. One of the advantages of such calibrated item bank is estimating the individual's ability based on calibrated tests with specific goals to decrease efforts and time consumptions and to enhance tests security. Indeed, leverage traditional paper and pencil test taking process to Computerized Adaptive Testing.