Description
Checks from CFA loadings that each construct's items converge (CR, AVE ≥ 0.50) and that constructs are distinct (Fornell-Larcker criterion or HTMT < 0.85).
Before interpreting structural paths, the measurement model must show that indicators measure their construct well and that constructs are not redundant. With standardised loadings λi and error variances θi = 1 − λi², composite reliability CR = (Σλi)² / ((Σλi)² + Σθi) and average variance extracted AVE = Σλi² / (Σλi² + Σθi). Convergent validity requires AVE ≥ 0.50; discriminant validity requires each construct's √AVE to exceed its correlations with other constructs (Fornell-Larcker) or the heterotrait-monotrait ratio to be below 0.85. Vivek Gopinath's payments case reports Trust CR 0.87/AVE 0.69, Fraud Risk 0.82/0.61, Slowness 0.81/0.57 and Support 0.83/0.62, with Cronbach's α of 0.86 (TR) and 0.79 (SL), and drops indicators loading below 0.60 unless theory demands.
Attributes
- Discipline
- Survey & Psychometrics
- Loss type
- Latent
- Stage
- Evaluation
- Data
- Quantitative
- Effort
- Hours
Steps
- Run CFA and obtain standardised loadings for each construct.
- Compute θi = 1 − λi² for each indicator.
- Compute CR and AVE per construct; require AVE ≥ 0.50.
- Compute √AVE and compare with inter-construct correlations (Fornell-Larcker), or compute HTMT and require < 0.85.
- Drop or revise indicators with loadings < 0.60 unless theory demands them, and re-run.
- Report CR, AVE, α and the discriminant test in the measurement table.
Sources
- Vivek Gopinath (2025), DAGs, SEM & Mixed Methods in UX: Turning Insight into Evidence in Banking & Payments
- Campbell & Fiske, Convergent and discriminant validation by the multitrait-multimethod matrix (1959), Psychological Bulletin
- Eid & Diener (eds.), Handbook of Multimethod Measurement in Psychology (2006)
- Tenenbaum, G. & Eklund, R. (eds), Handbook of Sport Psychology, 3rd ed. (2007)
Also known as
- Composite reliability and AVE
- Construct validity checks
- MTMM
- Campbell-Fiske matrix
- Convergent and discriminant validity analysis
- Multitrait-Multimethod Matrix
- rho_A
Categories
- Analytics & Behaviour
- Survey & Psychometrics