Feature Topic

Whereas variable-centered approaches (e.g., multiple regression, CFA, SEM) assume that all individuals from a sample are drawn from a single population for which a single set of ‘‘averaged’’ parameters can be estimated, person-centered approaches (e.g., mixture models, latent profile analyses, latent class analyses, mixture regression, growth mixture analyses) relax this assumption and considers the possibility that the sample might include multiple subpopulations characterized by different sets of parameters. Person-centered approaches thus provide a very rich complement to traditional variable-centered methods, allowing researchers to model complex processes in a more heuristic way (Wang & Hanges, 2011). Such methods are powerful tools to model interactions between large set of variables in an effective way. They identify naturally occurring subpopulations and allow for further comparisons between them. The raising popularity of person-centered approaches brings an opportunity for a more in-depth understanding of work-related phenomena. However, it also presents multiple challenges. For example, variableand person-centered approaches are often framed as distinct and complementary approaches to research, yet this complementary nature has seldom been systematically explored. A key issue that has yet to be explored in research is the impact of key decisions made in preliminary variable-centered applications on later person-centered results. Another key challenge is to further explore the full flexibility of the generalized structural equation modeling (GSEM) framework, which provides a very high level of flexibility to person-centered analyses over and above their classical formulations (e.g., correlated uniquenesses, heteroscedasticity, parameter constraints). Indeed, very little research carefully attempts to understand the meaning, in terms of theoretical implication, of these more flexible specifications made possible by GSEM. Another unexplored critical issue is that most applications of person-centered methods have been limited to providing a description of the results obtained in a specific sample, with at most an eyeball comparison of solutions obtained in multiple samples. This leaves open the question of whether (and how) results obtained from different studies using person-centered methodologies may be replicated, compared, or aggregated in a more quantitative manner? Finally, although it has often been argued that person-centered approaches should support theory development, strategies for both exploratory and confirmatory applications are needed.