Common Latent Factors in AMOS: Assessing Sensitivity to CMB
A CLF adds an unmeasured method factor loading on multiple items. It may absorb method variance, but it can also absorb cross-loadings, omitted substantive structure and item-specific errors.
Use a CLF as sensitivity analysis: compare the substantive model with a method-factor model, examine loading/path/fit changes and inspect the solution. Do not diagnose CMB from a 0.20 loading difference or improved fit alone.
CMV: prevention and diagnosis have distinct roles
- Design
Sources, wording and question order
- Collection
Procedures and response context
- Diagnosis
Combine relevant evidence
- Reporting
State measures and limitations
Conceptual diagram. One test cannot establish that a dataset is entirely free of CMB.
Source: Lê Minh · Conceptual illustration, not software output.1. What does a CLF assume?
In AMOS, a Common Latent Factor is usually a factor without dedicated indicators that loads on all items. It assumes a shared source broadly affects the measures (Podsakoff et al., 2012; Richardson et al., 2009).
A single homogeneous method source rarely captures every mechanism. A CLF may be unidentified, produce hard-to-interpret positive/negative loadings or compete with substantive factors.
Models with more parameters often fit better. The improvement may reflect omitted substantive content or cross-loadings rather than method alone.
2. A cautious specification workflow
Estimate the substantive model
Check CFA results, loadings, fit and the solution before adding a CLF.
Add the method factor
Connect the CLF to items using the selected identification strategy; do not fix parameters arbitrarily.
Inspect the solution
Check convergence, negative variances, unusual loadings and correlations near 1.
Compare parameters
Inspect changes in loadings, factor correlations and paths, together with CIs.
Consider alternatives
Compare conclusions with a marker or measured method factor when available.
3. Avoid mechanical loading-difference cutoffs
The frequently repeated 0.20 standardized-loading difference is not a universal test. Loading changes depend on identification, scaling and the model.
More useful questions are whether substantive parameters and hypothesis conclusions change materially, method loadings have an interpretable pattern, and the model remains stable.
| Evidence | Question |
|---|---|
| Solution | Does the model converge to an admissible solution? |
| Method loadings | Is there an interpretable pattern? |
| Substantive loadings | Are reductions uniform or localized? |
| Paths | Do signs or CIs alter the main conclusions? |
| Fit | Does the improvement justify the added complexity? |
4. Reporting template
Reporting example in English
A CLF/UMLV was added as a sensitivity analysis for same-source data. The model [converged/did not converge]; fit changed from [ ] to [ ]. Substantive loadings changed by [range] and the main paths [ ]. Results were [stable/sensitive], but the CLF was not treated as the sole evidence for ruling out or establishing CMB.
- Validate the baseline model before the CLF.
- Describe method-factor identification.
- Check Heywood cases and convergence.
- Do not rely on one loading-difference cutoff.
- Compare main coefficients and CIs.
- Combine results with design evidence.
Conclusion
A CLF is useful as a stress test of conclusions, not a CMB detector. Its value lies in parameter changes and the plausibility of the method model.
References
- Podsakoff, P. M., MacKenzie, S. B., & Podsakoff, N. P. (2012). Sources of method bias in social science research and recommendations on how to control it. Annual Review of Psychology, 63, 539–569.https://doi.org/10.1146/annurev-psych-120710-100452
- Richardson, H. A., Simmering, M. J., & Sturman, M. C. (2009). A tale of three perspectives. Organizational Research Methods, 12(4), 762–800.https://doi.org/10.1177/1094428109332834
Methodological note: If the method factor has no indicators measuring a specific mechanism, interpretation should remain diagnostic rather than naming it as an established cause.