Exploration On Item Selection Methods In Computerized Adaptive Testing

In the Computer Adaptive Testing, the item selection has always been our greatest concern. We not only want to examinee's ability to make an accurate estimate, but also want to make the item bank where the item is fully rational use of adaptive test item bank to improve security. In a traditional Computer Adaptive Testing, we tend to rely on the maximum information(MI) approach to select the item, although the MI approach is very efficient to estimate the examinee's ability, but the MI approach would lead to exposure of some of items imbalance and the high rate of item overlap, so that the security of the examination posed a certain threat. In order to reduce the visibility of the project reducing item overlap. And the MI approach in comparison methods, this paper will use a weighted likelihood estimation method of simulation of the maximum information approach, multiple stratification(MS) approach and mixture item selection approach(MS-MI ), these methods are the ability to accurately estimate and control item's exposure on the have certain advantages.This article by computer simulation method is to compare several methods described above.