Goodness-of-fit Item Test And Its Application Under Cognitive Diagnosis Framework

Cognitive diagnostic test is more intensive knowledge or skills test. In order to get effective and accurate diagnosis and to exploit the advantages to the full, one very important point is must ensure that the model and data fitting. This paper carried out the study of the three: The first study tried to corporate these goodness of fit test statistics that popular and well-done under item response theory into cognitive diagnosis theory; The second study compared the feasibility and effectiveness of the goodness of fit test statistics in the different factors like the sample size, the test length, the number of cognitive attributes; The third study analysised the real data to detect the application in real work. The findings showed:1) TheⅠand Ⅱ type error ration of these statistics(2?、2G、 2M、eq,D?、eq,GD) were less than 5%,which indicated that they could effectively detect the misfit items in cognitive diagnosis.2) From low to high principle, theⅠtype error ration of these statistics were eq,D?、eq,GD 、2?、2G、 2M and theⅡtype error ration of these statistics were 2?、2G、 2M、eq,D?、eq,GD.3) These statistics would be affected by the sample size, the test length, the number of cognitive attributes. Along with the growing number of the sample size, theⅠtype error ration of 2?、2G、eq,D?and eq,GD increased while the II type error ration of eq,D?and eq,GD decreased. Along with the growing number of the test length, theⅠtype error ration of these statistics(2?,2G, 2M,eq,D?,eq,GD) decreased but the II type error ration of them had no clearly change. Along with the growing number of cognitive attributes, theⅠtype error ration of eq,D?and eq,GD decreased while 2?and2G increased, the II type error of eq,D?and eq,GD increased while2?and2G decreased.4) When there had misfit items whose attributes were incorrectly calibrated in the cognitive diagnosis test, these statistics(2?、2G 、 2M、eq,D?、eq,GD) could effectively find out them.5) The eq,D?and eq,GD statistics worked better than the 2?、2G statistics in the real data.In a word, these statistics(2?、2G 、 2M、eq,D?、eq,GD) could be used on goodness of fit test in cognitive diagnosis theory.