A Sequential Method Based On Response Time For Detecting Compromised Items In Computerized Adative Testing

In recent years,with the development of computer network technology,computerized adaptive testing has been applied universally.Security issues of item pool of computerized adaptive testing attract attention of many scholars.Computerized adaptive testing is different from traditional pen-and-paper test,it can select a unique and individual test for each examinee based on the actual situation and test rules,thus it can improve test accuracy.But there are some problems in the computerized adaptive testing,such as items compromised,i.e.items leakage.Zhang(2014)has proposed a sequential method based on item response variable for detecting compromised items,and confirmed the effectiveness of monitored results.With the development of computer technology,the item response time of examinees can be accurately recorded.Therefore,in this paper,we can use these response time data to explore whether items are leaked in computerized adaptive testing.The specific method is to build a series of statistical hypothesis testing procedures based on item response time,and use sequential method to detect whether the items have been leaked when the test is running,and then demonstrate the effectiveness of this detecting method by simulation studies.At the same time,we compare this method with sequential method of detection based on item response variable.Simulation results show that: The sequential method of detection based on item response time has a bigger rate of Type I errors than that based on item response variable,but comparatively,the former has a bigger power value,a smaller lag and difference of detecting location.Therefore,the sequential method of detection based on item response time has higher sensitivity and useability.Additionally,this paper also studies the relationship between moving sample sizes with lags and difference of detecting location in two kinds of sequential methods of detection,so as to provide references for the construction of a more effective monitoring method.