Confirmatory factor analysis
Test a specified relationship between factors and indicators
Confirmatory factor analysis specifies how observed indicators relate to latent factors, including fixed zero loadings and justified residual relationships. It tests implications of a proposed measurement structure rather than discovering one without constraints. Estimator choice depends on indicator scale and distribution. Comparing groups or occasions may require measurement-invariance assessment.
Use when theory or prior exploratory work specifies an interpretable measurement model. Plan identification, indicator quality and adequate precision, preferably using data independent of the original structure discovery.
Strengths
- Makes measurement assumptions explicit and testable
- Allows latent-variable comparisons with modelled measurement error
Limitations
- Global fit can mask local misspecification
- Post-hoc modifications risk fitting sample-specific noise
Know the boundary
Good fit does not prove a factor structure unique or establish causal measurement theory.