“熵增值”在试题质量分析中的应用

Quality Analysis of Items Based on Increased Value of Entropy

难度和区分度是传统试题质量分析所采用的指标,而依据(难度,区分度)二维向量指标对试题质量进行排序需要主观权衡,从而造成了试题质量评判的不稳定性。基于此,根据信息熵理论,构建一种基于考生得分分布变化的客观试题质量指标"熵增值"。通过HSK阅读分测验的实证分析,"熵增值"的大小有效地反映了试题质量的优劣,其对应的难度和区分度指标符合经典测量理论的分析原则。最后采用模拟仿真的方式论证了"熵增值"和难度、区分度的相互影响关系。

Difficulty and discrimination are traditional index in item analysis. To distinguish the quality of items and sequence the items basing on difficulty and discrimination need subjective weigh, so the instability of the items estimation cannot be avoided. So according to the information entropy theory, increased value of entropy is constructed as the index to measure the quality of items basing on the examinee score distribution. Through the empirical analysis of HSK reading test, the entropy value reflects the quality of items effectively; and they are consistent with the indexes in Classical Test Theory. In the end, It is showed that increased value of entropy is affected by difficulty and discrimination by using analog simulation.