摘要
对"增大被试群体异质性,能提高测验信度"观点提出质疑。讨论非正态分布测量数据的偏度对克隆巴赫信度α系数的影响,以模拟方法验证Box-cox正态化变换对信度的高估现象,进而给出对克隆巴赫信度α系数估计的改进方法。
In order to challenge the notion of ＇ increasing heterogeneity of the subjects can result in higher credibility and reliability of the test＇ , this paper mainly discusses the effects of non-normal distribution measurement data＇s skewness on Cronbaeh＇s alpha reliability coefficient. The influences of Box-Cox normal-transformation on the estimation of reliability are analyzed by simulation and a method to improve Cronbach＇s alpha reliability coefficient is suggested.
出处
《应用心理学》
CSSCI
2008年第3期276-281,共6页
Chinese Journal of Applied Psychology
基金
广东省普通高校人文社会科学“十五”规划研究重点项目(06JDXMXLX002)
关键词
非正态分布
被试群体异质性
克隆巴赫α
Box—cox变换
non-normal distribution, heterogeneity of the subjects, Cronbach＇s alpha reliability coefficient, box-cox normal-transformation
