四参数模型对被试作答异常现象的拟合与纠正

4PM Rectifying the Overestimation and Underestimation in the Tests

心理与教育测验中存在着被试作答异常现象(能力测验中的猜测现象和睡眠现象,人格测验中的非0下渐近线现象和非1上渐近线现象),会导致被试能力或人格特征的测量偏差。在能力测验中,研究者已提出了多种方法来纠正猜测现象和睡眠现象,这些方法往往需要调整或删除被试作答信息,而四参数模型不需要改变被试作答信息而能有效纠正被试能力高估或低估现象。在人格测验中存在着非0下渐近线和非1上渐近线现象,四参数模型能增强测验项目拟合性能,提高人格测验的准确性。

In the educational and psychological tests, there exist guessing phenomena that the low-ability examinee makes correct answers on the difficult items, and the sleeping phenomena that the high-ability examinee makes wrong answers on easy items, which would lead to errors. Many researches have been done on the issue of these phenomena, and various methods have been proposed to rectify the overestimation and underestimation. Usually when these methods are used, these responses on items that appear to be too easy or difficult for the examinee need to be down-weighted or omitted. However, responses of the examinees need not to be changed with the four parameter logistic model in which c parameter reflects the guessing phenomena and rectifies the overestimation, γ parameter reflects the sleeping phenomena and rectifies the underestimation. In the personality testes, there exist some items which have non-zero lower asymeptote or non-one upper asymeptote. Four-parameter model could better fit the response data for the test and raise accuracy.