| name | questionnaire-sem-sop |
| description | Governed workflow for questionnaire-based CFA and SEM with an R/lavaan-first execution path. |
Questionnaire SEM SOP
Purpose
This project-level skill captures a reusable workflow in the survey domain.
Use when
Use when a questionnaire project needs CFA, SEM, indirect effects, or multi-group invariance planning before a stable plugin contract exists.
Inputs expected
- Scored questionnaire dataset or scoring-ready dataset plus codebook
- Explicit construct definitions and indicator-to-factor mapping
Source materials
../docs/questionnaire-sem-notes.md (sha256: 4e9b920bdf4d9b20fb0ff4e1598d5cbdcd8104d5354fed688305d078691e79e4)../docs/research-method-standards.md (sha256: 8223885b1d76ee9c65da2fa3f3df0637957aece16f8fe723be303844a480d812)../docs/project-skill-synthesis.md (sha256: 6625f3ee755659b62c4c341d5c060c241af25a14649432c89ac37e26b1098a3f)
Workflow
- Confirm construct definitions, scoring direction, reverse-key handling, and missing-data policy before modeling.
- Choose observed vs latent framing explicitly, and default SEM fitting to the R/lavaan path.
- Fit CFA before structural claims when latent constructs are central, then document fit indices, loadings, and warnings.
- For multi-group requests, record an invariance plan before comparing latent means or structural paths.
Expected outputs
- Model specification note
- Fit summary with warnings and standardized effects
- Interpretation checklist for CFA/SEM/invariance conclusions
Limits
- This skill does not replace statistical review, publication judgment, or advanced SEM diagnostics.
Failure checks
- Stop if construct-to-item mapping is ambiguous or reverse-key handling is unverified.
- Stop if the project requests group comparisons without a measurement invariance plan.
Evaluation examples
- Primary project example: stress, coping, and burnout modeled with latent CFA + SEM paths.
- Held-out example: compare configural, metric, and scalar invariance across gender groups before interpreting structural differences.
Governance metadata
- generated_at:
2026-04-03T02:48:14Z
- maturity:
project
- verification:
held-out-validated
- held_out_validation:
passed
Maintainer notes
- Keep interpretation guidance separate from executable tool contracts.
- Refresh source material provenance when the workflow changes materially.
- Promote stable computation into an external or bundled plugin only after the computation boundary stabilizes.
