Abstract

Invariance of surveys across different groups means that the respondents interpret the items in the same way, as reflected in similar factor loadings, for example. Invariance can be assessed using various statistical procedures, such as Multi-Group Confirmatory Factor Analysis. However, these analyses require access to raw data. Here, we introduce a meta-analytic method that requires only the factor correlation matrices of samples as input. It compares the structures of intercorrelations of factors by correlating these values across two samples, yielding a value of overall similarity for how the factors intercorrelate in different samples. This method was tested in three different ways. We conclude that the method yields useful results and can assess invariance when raw data are not available.

摘要

不同组间调查的不变性意味着，受访者以相同的方式解释项目，例如反映在相似的因素负荷中。可以使用各种统计程序来评估不变性，例如多组确认性因子分析。但是，这些分析需要访问原始数据。在这里，我们介绍一种仅需将样本的因子相关矩阵作为输入的元分析方法。它通过将两个样本中的这些值相关联来比较因素互相关的结构，从而得出各个样本中因素如何互相关的总体相似性值。该方法已通过三种不同方式进行了测试。我们得出结论，该方法产生有用的结果，并且在没有原始数据时可以评估不变性。