A Comparison Of Concurrent Calibration And Separate Calibration Under Neat Design Of Irt

Test equating is based on examination of the same kind of psychological qualitytest of multiple forms to transform score system,so that test scores of different testforms can be compared. Many contemporary large-scale tests use different forms ofthe same test, so how can we to compare with different examinee’s score in the exam.And in education measurement, with students of grade rise, how to depict the statusand longitudinal trend of development of the students’ ability and how to compareperformance between different age paragraph or grade or year, which depend on theequivalent technology to resolve the problems. Then it inevitably involves the choiceof what kind of the equivalent ways, so it is particularly important to comparebetween the equivalent ways.Under the framework of item response theory, there are two ways to equate: oneis separate calibration, it firstly estimates the items parameters of each test. As theindeterminacy of scale in IRT, the relationship between the parameters of differenttests is linear, all the item parameters will on a common metric after the lineartransformation. The main way of transformation: MM,MS, Haebara, Stocking-Lord,Another is called concurrent calibration, all the item parameters are estimatedone-time.In previous studies, the research on concurrent calibration and separatecalibration had the following defects and problems: First, the indicators of comparisonis not unified, different research used different index that may affect the results of thestudy. Secondly, just from the error value in the study to determine which is better, itdoesn’t explain what method is good or bad from the statistical.Against the above, in this study, at first, absolute deviation and the deviation isused to study of the random error and systematic error to ensure the accuracy of theresults. Then using a statistical test method to examine whether there is a significantdifference in statistics between concurrent calibration and separate calibration.According to the results of this study, We concluded that: First,On a parameter,equivalent effect of concurrent calibration is significantly better than that of separatecalibration. On b parameter, from the angle of deviation, Haebara method of equivalent effect is significantly superior to the other four, namely the separatecalibration significantly better than concurrent calibration. From the perspective of theabsolute deviation, there are no significantly difference in the equivalent effect ofconcurrent calibration, SL and Haebara method. Secondly, the larger of the samplesize is, the better of the equivalent effect will be.