论文部分内容阅读
目的:考察样本中的不拟合被试对CTT框架下测验信度、效度和IRT框架下测验信息量的影响。方法:使用lz指标和数据净化的方法,比较不拟合被试删除前后的分半信度、α系数、测验信息量的变化以及验证性因素分析的效果。结果:①随着删除的不拟合被试比率的增加,E量表与N量表的分半信度、α系数提高;②删除不拟合被试后,E、N量表对两因素验证性因素分析模型拟合更好;③不拟合被试的删除,可以提高测验信息量,降低测量标准误。结论:样本中不拟合拟被试的存在,会造成对测验信度系数的低估,影响测验的结构。从IRT的角度而言,会造成测验信息量的降低。
OBJECTIVE: To investigate the influence of non-fitting samples on test reliability, validity and test information under the framework of CTT under the framework of CTT. Methods: Using the method of lz index and data purification, the scores of half-confidence, alpha coefficient, the amount of test information and the effect of confirmatory factor analysis before and after the test were not fitted were compared. Results: (1) With the increase of the proportion of non-fit test subjects, the semi-reliability and the α coefficient of E and N scales increased; (2) After deleting the non-fitting subjects, Confirmatory factor analysis model fitting better; ③ not fit the subjects to delete, can improve the test information, reduce the measurement error. Conclusion: The existence of non-fit test subjects in the sample may lead to an underestimation of the test reliability coefficient and affect the structure of the test. From the IRT point of view, will result in reduced test information.