Structural equation models

Structural Equation Models (SEMs) constitute a family of parametric statistical models able to structurally relate a number of latent constructs that are indirectly observed through a set of indicators. A SEM consists of two parts: the measurement part, which defines the relationship between each construct and its indicators through a linear factor model, and the structural part, which defines the structural relationships among the latent constructs through a system of linear regressions. Both the measurement and the structural part of a SEM admit a large variety of specifications allowing to test complex structural hypotheses involving the latent constructs. SEMs are widely employed to analyse both observational and experimental data in economic, social and psychological sciences.