Skip to content

Replication & Cases

Risk Bridge includes four curated, privacy-safe replication case studies packaged directly in the distribution. These harnesses verify statistical validity, numerical stability, and performance scaling without exposing sensitive patient data.

Available Cases

  • Reproduction Runbook: Master runbook for regenerating all public artifacts.
  • Numerical Validation Suite: Compares analytic gradients against finite-difference approximations, evaluates optimizer trajectories, and verifies parameter recovery.
  • External Calibration Validation: Monte Carlo evaluation (\(N_{\text{sim}} = 50\)) testing the recovery of external calibration benchmarks under matched and degraded conditions.
  • Synthetic Transport Case: End-to-end simulation of transport between disparate populations with covariate shift and selection bias.
  • Runtime Support Scaling: Evaluates computational runtime and memory consumption across expanding Cartesian covariate supports.

Running Cases via Python Modules

All cases can be executed directly as Python modules from a standard install or source checkout:

# 1. Numerical validation
python -m cases.numerical_validation.run_suite

# 2. External calibration smoke run
python -m cases.external_calibration_validation.run_suite --profile smoke --condition matched

# 3. Synthetic transport example
python -m cases.synthetic_transport_example.run_case

# 4. Runtime scaling smoke profile
python -m cases.runtime_support_scaling.run_scaling --profile smoke