Stress-Test Families
These tables specify controlled perturbations and their expected diagnostics. They are designed to be used as research appendices without requiring interactivity.
| ID | Test | Perturbation | Range | Primary metrics | Interpretation risk |
|---|---|---|---|---|---|
| T1 | Symmetric noise | D, k | ±10%, ±25% | flip count; Δ drift; near-boundary time | Hidden fragility if |λ| small |
| T2 | Asymmetric shock | D ↑ (k fixed) | +15% → +50% | time-to-flip; robustness erosion | Underreach if posture unchanged |
| T3 | Over-control scan | k ↑ above boundary | +25% → +150% | stability invariant; proportionality check | Overreach risk despite stability |
| T4 | Boundary drift | k* drift | slow drift + bursts | near-boundary occupancy; oscillations | Decision oscillation; governance confusion |
| T5 | Coefficient uncertainty | α, β | ±5% → ±30% | confidence bands for λ, Δ | False certainty in outputs |
How to read these tests
A robust framework shows invariance under mild perturbations and clearly flags near-boundary regimes where
small drift can flip decisions. FDRSM-5 is the documentation of those safe/unsafe regions.