The Analysis of Mean Differences Using Mean and Covariance Structure Analysis
展示了在比较组间均值时，如果测量工具不具备测量不变性，可能会夸大、缩小甚至颠倒均值差异的方向，强调了忽视测量不变性的统计和实质后果。
Many studies in the social and organizational sciences are concerned with estimating and testing between-group mean differences. Typically, responses to these scales are summed, and the summed score or the mean of the summed score is then compared across groups using standard statistical tests ( t tests) and described using d values. Even though there is a call for establishing measurement invariance prior to examining mean differences, very few studies seem to follow this advice. In this study, the authors extend prior invariance work to demonstrate how a lack of invariance can obscure the effect size, direction, and statistical significance of mean differences. In particular, the authors show how manifest mean differences can be exaggerated, reduced, or even switch in direction relative to latent mean differences. Thus, they emphasize the statistical and substantive consequences that may occur when the scales used to perform mean comparisons do not possess measurement invariance.