ABSTRACT
PURPOSE: In order to apply the item response theory to test results, the assumptions made about unidimensionality and item goodness-of-fitness should be tested before analysis. To know if the Korean Medical Licensing Examination results fulfill these assumptions, appropriate tests were done on these two features. METHODS: Three results(from three subjects) of the Korean Medical Licensing Examination, done in January 2004, were converted into 1, 0 data. The Holland and Rosenbaum Method(HRMH) and DETECT were used to test unidimensionality. The Winsteps was used to test goodness-of-fitness according to the Rasch model. RESULTS: Unidimensionality was rejected by HRMH but accepted by DETECT. With the Rasch model, 2 out of 550 items were inadequate. Each 137, 73 and 455 examinees out of 3,881 were inadequate in three subjects of 126, 400 and 24 items. CONCLUSION: The above results suggest that DETECT is desirable for testing unidimensionality of high-stakes tests with more than 100 items. According to the Rasch model, the Korean Medical Licensing Examination fulfills the goodness-of-fitness to be analyzed according to the item response theory.