From implicit assumptions to explicit testing: Measuring constructs within a multilevel framework and evaluating their psychometric isomorphism

Researchers who use a multilevel framework often aggregate scores measured at Level 1 in order to measure the construct at Level 2. In doing so, they implicitly assume that the construct is measured at both the individual and cluster level, that the ontology of the construct is the same at both levels, and that reliability is good at both levels. The aim of this paper is to present a seven-step approach that allows one to explicitly test those assumptions. Multilevel confirmatory factor analysis, a statistical technique that control the nonindependence of data in measurement models testing a factor structure at the within-level and between-level simultaneously, provides an analytic framework in which it is possible to test psycho-metric isomorphism and reliability. Recommendations about what to do when assumptions are verified or not are provided in the Section “Discussion.” The Appendix reports Mplus syntaxes necessary to run each step of the analysis.