Labouvie E, Ruetsch C
Multivariate Behav Res. 1995 Jan 1;30(1):63-76. doi: 10.1207/s15327906mbr3001_4.
Most quantitative comparisons of interest are based on scores obtained from multi-item scales. In order for such comparisons to be valid and meaningful, conventional thinking dictates that simple structure and metric invariance must be satisfied at the level of individual items. However, it is proposed here that the use of multi-item scales requires only that conditions of simple structure and metric invariance be satisfied at the scale level, that is, for hypothesized sets of items as a whole rather than each item individually. The proposed concepts can be implemented within the context of confirmatory factor analysis (CFA) by imposing constraints on mean factor loadings and intercepts of item sets rather than loadings and intercepts of each individual item. Further, the present method is proposed as a conceptual alternative to the specification of measurement models that is currently predominant in CFA.
大多数令人感兴趣的定量比较都是基于从多项目量表获得的分数。为了使这种比较有效且有意义，传统观念认为在单个项目层面必须满足简单结构和度量不变性。然而，本文提出使用多项目量表仅要求在量表层面满足简单结构和度量不变性条件，也就是说，对于假设的项目集整体而非每个单独的项目。通过对项目集的平均因子载荷和截距而非每个单独项目的载荷和截距施加约束，所提出的概念可以在验证性因子分析（CFA）的框架内得以实现。此外，本文提出的方法是作为当前CFA中占主导地位的测量模型设定的一种概念性替代方法。
