Multilevel Latent Profile Analyses: A Comprehensive Guide

Abstract

The organizational sciences have recently seen an uptake in the use of both person-centered analyses, most typically implemented via latent profile analyses (LPA), and multilevel analyses. Yet their combination remains conceptually underdeveloped and forms the focus of the current article. On the one hand, LPA is perfectly suited to the identification of subpopulations, or profiles, of participants characterized by distinct configurations on a set of variables. On the other hand, multilevel models are aptly suited to uncover, or control for, phenomena unfolding across individual, group, and organizational levels. In combination, multilevel-LPA provides the best of both worlds, enabling researchers to identify subpopulations of individuals and/or groups corresponding to distinct prototypical states, while modeling within-group and between-group variability. In the current article, we provide a comprehensive guide for multilevel-LPA implementation to monitor group contexts, climates, and cultures. We present six types of multilevel-LPA specifications—additive, dispersion, additive-dispersion, dispersion-heterogeneity, restrictive cross-level, and full cross-level—that ought to be carefully selected based on the nature of the constructs being measured and the research questions being asked. In doing so, we aim provide researchers with the analytical and conceptual tools to select and apply the appropriate multilevel person-centered model to achieve their research objectives.