能力水平分组对Mantel-Haenszel方法检验DIF效应的影响分析

摘要

Mantel—Haenszel方法（以下简称M-H方法）是探测试题是否存在DIF现象的一类重要的和普遍的方法。能力分组是M—H方法应用的一个关键环节。本文以某年度6000名考生普通高考文科综合选择题的作答数据为样本，探讨了能力水平分组对该方法检验敏感性的影响程度。研究结果表明：不同能力水平分组的检验结果均具有较好的一致性i检验结果对能力水平分组组数的敏感性较小；M—H方法具有较好的稳定性。%Mantel--Haenszel is an important and general way of finding out if there is DIF in exam questions. Ability grouping is a crucial stage in the application of M-H. This study deals with the extent to which ability grouping affects the testing sensitivity of M-H, based on how 6000 test takers answered the comprehensive multiple choice questions for arts majors in a college entrance examination. The findings indicate that testing results by means of ability grouping have relatively great consistency, while their sensitivity to the grades of different ability groups is relatively small. M-H shows quite good stability.