Structural equation modelling
Connect measurement models with structural hypotheses
Structural equation modelling combines equations among observed or latent variables, often linking a measurement model to structural paths. Covariance-based SEM evaluates restrictions implied by the whole model. Path analysis is the observed-variable special case. Model identification, estimator choice and temporal or causal assumptions must be separated from the act of fitting arrows.
Use to evaluate a theoretically specified system of relationships where measurement error or several connected equations matter. Ensure identification and data quality, and justify any causal interpretation through design and assumptions.
Strengths
- Models measurement and structural relationships together
- Enables explicit tests of constrained theoretical models
Limitations
- Equivalent models may fit the same covariance matrix
- Complexity can exceed the information available
Know the boundary
An arrow in SEM does not establish causality; cross-sectional fit rarely identifies temporal mechanisms.