Development And Validation Of An Item Bank In Adaptive Testing

With the development of Chinese education informationalized construction and the advent of the "Internet +" evaluation era,large-scale rational evaluation combined with computer technology and measurement theory such as big data and Internet will become the trend in the future.At present,the public course of Modern Educational Technology still uses paper and pencil tests as examination form which has various shortcomings.If computer-based testing is applied,especially Computerized Adaptive Testing(CAT),we can make the examination more scientific and standardized.In order to carry out CAT,it is necessary to build a standardized item bank based on Item Response Theory.Firstly,this paper discusses Item Response Theory which is the theoretical basis of CAT,and introduces the hypothesis and parameter model of it.In order to construct scientific and standardized questions,the basis for making cognitive objectives of test questions,Bloom's taxonomy of educational objectives,and combines it with Modern Educational Technology is discussed.Finally,the content standard of the course is constructed according to the Standards of Information-based Teaching Ability for Normal Students,the National Educational Technology Program of the United States and the Educational Technology Ability Standards for Teachers of the United States.Cognitive criteria and content criteria are combined to form assessment criteria.Then the design and proposition of the test are carried out,and the standard-based test is constructed.After compiling 248 questions,four sets of parallel tests were designed with anchor test,and 1633 students were tested in the final examination.Therefore,an item bank system based on B/S is constructed,which can store questions and conduct online examination.Due to the large number of students per test,it is necessary to optimize the server and database configuration in order to solve the high concurrency problem.Finally,Item Response Theory is used to test the validity of item bank,parameter estimation and fitness test,test equating,differential item functioning and test information function to construct the final item bank.Under different test termination conditions,simulated CAT is carried out to test the efficiency of CAT.Finally,the following conclusions are drawn:According to the validity test of the results,the reliability and validity of the questions can meet the requirements and the two-parameter logistic model is the most suitable one.The anchor items are used to connect each set of tests,and the test linking and equating is carried out.After removing the non-fit and too difficult or too easy items,211 items are finally generated and put into the item bank.Statistical test information of test questions in question bank proves that the question bank can provide enough information for computer CAT.Subsequently,the item function differences of these questions were tested,and it was found that there was no difference in item function between male and female students.Using this question bank to implement simulated CAT,the efficiency of the item bank is analyzed under different test termination conditions.When the standard is set to 0.38,the accuracy rate is higher,and the average number of items used is 46,which can meet the requirements of the examination.However,the item bank has low efficiency in analyzing high-competence students.