bam-engine 0.4.0

This release adds the robustness analysis package (Section 3.10 of Delli Gatti et al., 2011), overhauls the calibration package, re-calibrates default parameter values, and fixes several bugs in internal event logic.

New: Robustness Analysis (Section 3.10)

• validation/robustness/ — internal validity (20-seed stability), sensitivity analysis (5 parameter experiments), structural experiments (PA removal, entry mechanism)
• New consumer_matching config parameter and extensions/taxation/ for structural experiments
• CLI: python -m validation.robustness

Calibration Package Overhaul

• Rebuilt into focused modules with Morris Method screening (default), tiered stability testing, and pairwise interaction analysis
• Phase-based CLI with --phase, --resume, --rank-by, --grid, markdown reports

Packaging

• extensions/, validation/, calibration/ are now pip-installable sibling packages

Bug fixes (event logic)

• Bad debt recovery/loss inversion, R&D deduction ordering, net_worth_ratio dimensional fix, multi-lender loans, spurious wage_bill recalculation, projected_fragility initialization, and more

Breaking changes:

• Re-calibrated defaults: 300→100 firms, 3000→500 households, new-firm entry parameters, job_search_method → "all_firms"
• Config removed: contract_poisson_mean, loan_priority_method, firing_method, matching_method
• destroyed → collapsed rename
• net_worth_ratio semantics changed to "multiple of initial revenue" (default 3.0→6.0)

See CHANGELOG for complete details of what's included in this version.

If you want to get started with using BAM Engine, see the README.