大样本条件下概率阈限的改变对参数估计的影响  

The Influence of the Change of Probability Threshold on Parameter Estimation under the Condition of Large Sample

作　　者：王风 Wang Feng(Institute of Language Testing and Human Resource Assessment,Beijing Language and Culture University,Beijing,100083)

机构地区：[1]北京语言大学语言测试和人才测评研究所,北京100083

出　　处：《考试研究》2021年第1期65-69,共5页Examinations Research

摘　　要：考生猜测作答是一种偏离测验构念的行为,可以通过ARRG法检验,并将与其相关的问题最小化。通过比较大样本情况下概率阈限的改变对参数估计的影响,发现ARRG法能明显提高数据-模型的拟合优度。大样本条件下,概率阈限的改变主要引起题目难度估计的变化,标准误的波动较小。本文为消除考生猜测作答的影响及使用ARRG法时根据样本量合理解释参数变化的意义提供参考。Guessing is a behavior that deviates from the test structure,which can be tested by the ARRG method and minimized as for the problems related to it.By comparing the influence of the change of the probability threshold on the parameter estimation in the case of large samples,it is found that the ARRG method can significantly improve the goodness of data-model fitting.Under the large sample condition,the change of the probability threshold mainly causes the change of the difficulty estimate of the problem,and the fluctuation of the standard error is small.This article provides a reference for eliminating the influence of guessing,and for explaining the significance of parameter changes according to the sample size when using the ARRG method.

关 键 词：RASCH模型 猜测作答 概率阈限 样本量 

分 类 号：G424.74[文化科学—课程与教学论]