ABSTRACT
This chapter considers the ways in which structural equation modeling (SEM) methods can be integrated with multilevel regression to investigate a wide variety of models. The model contains hierarchical data structures, focusing in particular on incorporating measurement error in defining constructs through their observed indicators. It first develops a single-level model with two latent factors. Then it extends this model to consider the nesting of individuals within departments. Finally, it extends this model to consider the nesting of departments within organizations. Latent variables can be used to represent a number of different statistical concepts including random coefficients, sources of variation in multilevel analyses, growth trajectories and latent classes. Specifying the measurement model involves defining the observed indicators in terms of latent factors and the technique used to define measurement models is referred as confirmatory factor analysis (CFA). It is referred to as CFA because of the emphasis on proposing a set of theoretical relationships to determine the proposed factor model.
