ABSTRACT
Making valid comparisons of latent variable (LV) mean scores across groups is anything but a trivial task as it relies on stringent measurement invariance conditions. This chapter presents a simulation study that investigates the extent to which noninvariance of factor loadings and indicator intercepts may lead to false statistical con- clusions in terms of LV mean differences across groups. In this simulation study, noninvariance conditions were represented by just one LV indicator with a noninvariant factor loading applied to the LV indicator and/or a noninvariant indicator intercept. The chapter provides measurement invariance conditions in addition to noninvariance conditions. The measurement noninvariance design factors determine the degree of measurement noninvariance of the noninvariant indicator. The simulation study has shown that a noninvariant LV indicator may have a very strong impact on the percentage of correct statistical conclusions of a LV mean difference test.
