Pokrajac N, Culo F
Department of Physiology, Faculty of Medicine, University of Zagreb, Croatia.
Med Educ. 1994 Sep;28(5):409-17. doi: 10.1111/j.1365-2923.1994.tb02552.x.
At the mid-term test in Part I Physiology at the University of Zagreb the students (n = 280) were graded by our standard pass level (SPL) arbitrarily set at 54% correct answers (SPL = 0.54). The test consisted of 50 items of the one best answer type. Items were selected from the pool by one examiner to conform, by his judgement, to the predetermined SPL. Post hoc the minimum pass level (MPL) was assessed independently by eight examiners and an MPL value of 0.60 for the whole test was obtained. The original Nedelsky scale was used in assessment of MPL but for statistical analysis the data was expressed as log(1/MPL) to linearize the scale of measurements and to reduce the variances. The data showed a large difference between examiners in their assessment of MPL. Nevertheless, the average log(1/MPL) value of individual items showed a significant negative linear relationship with the item difficulty indices as calculated from student's answers, indicating that despite the large heterogenity in assessment the average item log(1/MPL) may be acceptable as a reasonable prediction of item difficulty. Finally, 'subtests' were formed from the whole test by grouping items according to their log(1/MPL) value. The passing rate at these subtests was found to be identical despite the fact that they considerably differed in their MPL values. Therefore, the MPL value seems to be useful in setting objective standards for the decision of pass or fail, even when the MPL was assessed in a very heterogenous way.
在萨格勒布大学第一部分生理学的中期测试中，对学生（n = 280）按照我们任意设定为54%正确答案的标准及格水平（SPL = 0.54）进行评分。该测试由50道单项最佳答案类型的题目组成。题目由一名考官从题库中挑选，据其判断符合预定的SPL。事后，由八名考官独立评估最低及格水平（MPL），得出整个测试的MPL值为0.60。评估MPL时使用了原始的内德尔斯基量表，但为了进行统计分析，数据以log(1/MPL)表示，以使测量量表线性化并减少方差。数据显示考官在MPL评估上存在很大差异。然而，各个题目的平均log(1/MPL)值与根据学生答案计算出的题目难度指数呈现出显著的负线性关系，这表明尽管评估存在很大异质性，但平均题目log(1/MPL)作为题目难度的合理预测可能是可以接受的。最后，根据题目log(1/MPL)值将整个测试分组形成“子测试”。尽管这些子测试的MPL值有很大差异，但发现它们的及格率是相同的。因此，即使MPL是以非常异质的方式评估的，MPL值在设定通过或不通过的客观标准方面似乎也是有用的。
