Measurement Invariance (Time, Samples, Contexts)

Communication researchers regularly conduct comparisons between populations and comparisons over time. Typically, the objects of a comparison are relationships between constructs, such as media effects, or the means of constructs, such as average media use. Importantly, comparisons of relationships and means are based on the assumption that the involved constructs are measured the same way across populations or over time. If this assumption holds, the measurements are invariant across populations or over time. However, most comparative studies do not explicitly test for measurement invariance. This is problematic because results of comparisons can be biased when invariance assumptions are violated. Measurement invariance must thus be tested in all comparative communication studies. This entry discusses the different forms of measurement invariance for continuous data and how they can be tested in multigroup confirmatory factor analyses. Furthermore, partial measurement invariance, and invariance tests for ordinal data and for formative measurement models are discussed.