摘要
目的:用个人拟合指标对EPQ量表数据进行分析,探讨不拟合现象的可能成因和不拟合被试对参数估计的影响。方法:用lz和Guttman错误检查不拟合情况,用RMSD、MD指标及ICC面积法考察参数估计结果的差异。结果:①lz<-1.65的被试占总人数的3.44%,lz与Guttman错误的相关为-0.838;②最不拟合的50人中,编号相连被试共有32人;③删去不拟合被试前后,特质参数差异很小,高区分度高难度和高区分度低难度项目参数差异大。结论:个人拟合指标可以找到团体测验中来自同一测试环境的异常反应被试集团;不拟合被试的存在,对被试特质参数的估计影响不大,但会造成对项目区分的低估,进而影响难度越大和越小的项目。
Objective： Person-fit indices were computed on EPQ.This study was to investigate the cause of misfit and the influence of the inclusion of misfitting persons in parameter estimates.Methods： lz and Guttman error were used to investigate the misfit.MD,RMSD indices and ICC method was used to evaluate the effect of deletion of mifitting persons.Results： ①lz-1.65 was found in 3.44% of the examinees,and lz was correlated-0.838 with Guttman error;②There were 32 in 50 most misfitting persons ranked in adjacent numbers;③It was found that person parameters changed little after deleting misfitting persons.However,the high α high β items and the high α low β items changed a lot.Conclusion： A group of aberrant response persons in the same testing environment were found out by person-fit indices in group test.The presence of misfitting person cannot influence person parameters much,but resulted in underestimating discrimination parameters,and then influenced the bias in estimation of higher or lower threshold parameters.
出处
《中国临床心理学杂志》
CSSCI
CSCD
2011年第3期323-326,共4页
Chinese Journal of Clinical Psychology
