Governance
Model comparison and validation
Compare OLS, rolling windows, regularized models, and stress templates with transparent fit, stability, and interpretability tradeoffs.
Governance layerGuided intake
Start with a decision, not a blank prompt
The workshop guide uses these intake points to ask sharper questions before producing structured output.
Goal
What decision should model comparison support?
Inputs
Which dataset, ticker, scenario, or assumption set should be trusted?
Constraint
What risk limit, horizon, or caveat should shape the output?
Output contract
Every useful answer becomes a card stack
Recommendation
What the workflow suggests and why it is not automatic advice.
Evidence
The metrics, chart movement, or model signal supporting the view.
Caveat
The assumption, missing data, or model risk that could change the result.
Next step
The most useful follow-up action before exporting or sharing.
Best validation
Ridge
Demo ranking after stability penalty
Drift flag
Medium
Coefficient movement above monitoring threshold
Models tracked
4
OLS, rolling OLS, ridge, and scenario templates
Professional review checklist
- OKSeparate model-selection metrics from investment conclusions.
- OKTrack when training windows or factor definitions change.
- OKRequire human sign-off before operational use.
Empty state behavior
Run at least two model variants to populate the comparison matrix.
Correlation is not causation. Results are model-dependent and assume regime stability.
Placeholders only. No payment gate.