基于多层验证性因素分析的各种信度系数方法

近年来,心理学研究的复现性受到广泛关注,许多实证研究难以重复验证,信度较低。大量研究使用多层技术,但只报告整体信度,导致研究可重复危机。基于各种信度系数和验证性因素分析的理论,以二层模型为例,总结多层信度计算方法,通过文献综述检索应用情况,并用MPLUS进行实例演示,最后讨论单层信度估计存在的偏差及分层估计的好处。总之,对多层数据进行分层信度估计是很有必要的,可消除因测量工具缺乏信度而导致的研究不可重复。

Various Reliability Coefficient Methods Under Multilevel Confirmatory Factor Analysis

In recent years,the reproducibility of psychological science has received extensive attention.Many empirical studies were difficult to replicate and therefore had low reliability.It is noted that more and more researchers used a multilevel data structure and hierarchical linear modeling(HLM)for analysis.However,most of them only displayed the overall reliability,resulting in the research reproducibility crisis.Based on the confirmatory factor analysis(CFA),the article introduced estimation methods of different reliability coefficients.Taking a two-level model as an example,it also summarized the multilevel reliability estimation methods.In addition,to evaluate the application of multilevel reliability,the article investigated the researches in the past five years.Then,it illustrated how to use the software MPLUS to execute reliability estimation with an example.Finally,the article discussed the deviation of single-level reliability estimation and the benefits of hierarchical estimation.In summary,it is necessary to perform hierarchical reliability estimation in the multilevel data structure,which can eliminate the reproducibility crisis caused by unreliable measurement tools.