When to use this
Use before analysis and whenever a new dataset version or scoring decision changes the evidence base.
Why it matters
Transparent algorithms separate a review flag from a researcher-approved data treatment.
Step-by-step instructions
1. Run diagnostics on an immutable version.
2. Select relevant variables or item batteries.
3. Review flags in context.
4. Preview a transformation.
5. Approve only with justification to create a child version.
Expected outputs
Missingness, distribution and response-quality summaries; VIF; transformation preview; immutable version provenance.
How to interpret the results
Diagnostics identify evidence for review. They do not automatically define invalid data.
Common mistakes
Do not delete flags automatically, overwrite the source, or apply reverse coding without confirming scale endpoints.
Was this article helpful?
Your response helps the methodology team improve this documentation.