The Usability of Item-Total Correlation as the Index of Item Discrimination

Hwang, In Hong

Abstract

Item analysis is the evaluating process of items used for tests. Item difficulty, discrimination, and distractor analysis are the main components of the analysis. Discrimination index(D) by the method of extreme groups had been used for the item discrimination, but it had been known to have some disadvantages compared to item-total correlation(ITC). This study was conducted to evaluated the feasibility and the advantages of the ITC. Medical specialist qualifying examination carried out in Jan. 1999 was selected for the study material and the items of tests for the 4 major disciplines(internal medicine, general surgery, pediatrics, and obstetrics & gynecology) were analysed. The numbers of the items and examinee are 120 items/428 persons, 140/219, 140/229, and 140/226 (in the order of IM, GS, Ped, OB & Gyn) respectively. The average discrimination index(D) of all items is 0.170 and the standard deviation is 0.120. For the ITC, average is 0.210 and standard deviation is 0.117. There is positive correlation between D and ITC(r=0.677). The variation of the ITC is 0.880, which is wider than that of discrimination index(D), 0.713. Especially on the items with item's p-value greater than 0.9(n=140), the variations are 0.542 and 0.273 respectively. The difference is much distinct. These results imply that ITC can be used as the index of the item discrimination, and has some advantages compared to discrimination index(D). The advantages are the significance of the number itself and rather independence from the item difficulty.