Model Validation
Validation asks whether a model is adequate for a particular decision, not whether it is universally true. Fit, prediction, calibration, stress behavior, and operational use are distinct tests.
Central questions
- Which observable implications were not used to fit the model?
- How are parameter, sampling, and structural uncertainty reported?
- What decision loss results when the model is wrong?
Planned model
Compare models that fit the same in-sample statistic but disagree on tails, event dependence, fills, and policy outcomes under held-out regimes.