Statistics Glossary → Measurement invariance (configural, metric, scalar, strict)
Measurement invariance (configural, metric, scalar, strict)concept
Testing, with progressively stricter equality constraints, whether an instrument measures the same construct in the same way across groups (sex, country, time).
Invariance is tested step by step with multi-group confirmatory factor analysis. Configural invariance means the same factor structure holds in every group. Metric (weak) invariance means the factor loadings are equal across groups, which allows relationships between factors and other variables to be compared. Scalar (strong) invariance adds equal item intercepts; this is the level needed to compare latent means. Strict invariance adds equal residual variances and gives the strongest support for comparing observed total scores.
Each step is compared with the previous one through the chi-square difference and the change in fit indices; because chi-square is very sensitive in large samples, Chen's criteria (a CFI drop smaller than 0.01, an RMSEA rise smaller than 0.015) are widely used. If full invariance fails, a partial invariance model can be fitted in which a few problematic parameters are freed; that decision should have a theoretical justification.
An everyday example
A happiness scale is given in two countries and the second country scores higher. But the intercept of the item 'I smile often' is clearly higher in the second country: in that culture smiling is more common regardless of happiness. Because scalar invariance does not hold, part of the mean difference may come from measurement bias.
In MerQur
'Measurement invariance (group)' and 'Partial invariance (free parameters)' in Confirmatory Factor Analysis; the 'Group variable (multi-group)', 'Measurement invariance' and 'Partial invariance' options in Structural Equation Modeling; chi-square differences, CFI and RMSEA differences and score tests are reported.
Analyses where this concept appears
Confirmatory Factor Analysis (CFA)Structural Equation Modeling (SEM)
Related concepts
Measurement model and structural modelModel fit indices (χ², CFI, TLI, RMSEA, SRMR)