K-means聚类方法在中考标准设定中的信度分析  

Reliability Analysis of the K-means Clustering Method in Standard Setting for Junior High School Academic Proficiency Examinations

作　　者：温红博 刘先伟 姜有祥 WEN Hongbo;LIU Xianwei;JIANG Youxiang(Beijing Normal University,Beijing 100080,China)

机构地区：[1]北京师范大学中国基础教育质量监测协同创新中心,北京100080

出　　处：《中国考试》2024年第8期69-78,共10页journal of China Examinations

摘　　要：中考的标准设定是划分考生分数等级的统计技术,其质量高低关系到分数报告的可靠性和公平性。从我国东、中、西部地区各随机抽取3000名考生的中考数据,探讨K-means聚类方法在中考标准设定中的信度问题,从经典测量理论、概化理论和项目反应理论三个角度开展分析。结果显示,K-means方法在分数等级为四或五级时分类信度符合测量学标准,而对低分段的考生分类信度高于高分段的考生。总体来看,K-means方法适用于较低分数等级的标准设定中,可为中考标准设定提供分界分数的参考。The standard setting in junior high school academic proficiency examinations is a statistical technique for categorizing students'score levels.Its quality is vital for the reliability and fairness of score reporting.This study use data from 3,000 randomly selected students in junior high schools across the eastern,central,and western regions of China to investigate the reliability of the K-means clustering method in setting examination standards.The analysis,approached from classical test theory,generalizability theory,and item response theory,shows that the K-means clustering method has acceptable classification reliability when the number of score levels is four or five,with higher classification reliability for lower-scoring students.The findings suggest that the K-means clustering method is suitable for setting standards for lower score levels and can provide reference boundary scores for the standard setting of junior high school academic proficiency examinations.

关 键 词：中考 标准设定 K-means方法 聚类分析 

分 类 号：G405[文化科学—教育学原理]