In many research settings, variability, rather than means, is of key interest. For example, it may be critical for researchers to understand what factors have an impact on individual status… Click to show full abstract
In many research settings, variability, rather than means, is of key interest. For example, it may be critical for researchers to understand what factors have an impact on individual status stability or team cohesiveness, or how these variability-related characteristics would affect other outcomes. It may also be of interest for researchers to examine whether withinteam cohesiveness (or within-person stability) changes over time or differs between different types of teams (or individuals). Such research questions call for a modeling framework that can partition the observed variance into components at different levels of analysis. Within this multilevel modeling framework, the lower-level random effects variance should be allowed to vary across the higher-level units; importantly, this variability is of focal interest. We propose an analytical framework based on multilevel SEM (MSEM) to model the random effects in the lower-level variance components. Within MSEM measurement models can be incorporated at both lower and higher levels to account for measurement errors. Random effects can be modeled as latent variables; these random effects can be embedded within a more general latent variable framework and serve as outcomes and/or predictors. In the current study, two different modeling approaches, the logtransformation approach and the phantom variable approach (e.g., Stapleton, Yang, & Hancock, 2016) were investigated (Figure 1). Illustrative examples of unconditional models and conditional models with measured outcomes or latent outcomes were provided. Technical details of the model estimation were discussed together with exemplary interpretations of the model results. Our preliminary results suggested that the proposed analytical framework based on MSEM is highly
               
Click one of the above tabs to view related content.