New Ways of Dealing with Lacking Measurement Invariance

Abstract

AbstractIn educational research, comparisons are often made of groups or of the development of various (latent) constructs over time (e.g. teaching quality in different countries or different groups’ (girls vs. boys) perceptions of teaching quality). However, before the results of such comparisons can be accurately interpreted, measurement invariance (MI) of the constructs under investigation needs to be established to ensure their meaning remains consistent across groups, subjects, or assessment points. Thus, if mean level changes are to be compared between groups, scalar factorial invariance needs to be established. In this chapter, we investigate and discuss how results of MI analyses should be interpreted and whether they should be reported on with regard to contents. Using data from the well-known Programme for International Student Assessment (PISA) study on teaching quality, we introduce an approach to examining the conditions under which comparison among cultural groups is possible even if MI is lacking.