Methodology & Data Provenance

Complete documentation of 25 Python visualization scripts

Data Sources

panel_data.csv
Full Pathoutput/data/panel_data.csv
OriginBIS Data Portal
URLhttps://data.bis.org/static/bulk/WS_CREDIT_GAP_csv_flat.zip
Generated Bypipeline.py -> BISDataFetcher.get_panel_data()
MethodologyOne-sided HP filter with lambda=400,000
ReferenceDrehmann & Tsatsaronis (2014), BIS Quarterly Review
Rows3,719
signal_extraction.csv
Full Pathoutput/analysis/signal_extraction.csv
OriginCalculated from panel_data.csv + crisis_database.py
Generated Bypipeline.py -> CrisisPredictionModel.evaluate_thresholds()
MethodologyBinary classification: gap >= threshold signals crisis within 12 quarters
extended_results.json
Full Pathoutput/extended_results.json
OriginML model training on panel_data + crisis labels
Generated Byrun_extended_analysis.py -> MLCrisisPrediction
Methodology70/30 chronological split, 12 engineered features
crisis_database.py
Full Pathsrc/bis_research/data/crisis_database.py
OriginLaeven & Valencia (2020) + 2023 FDIC/FINMA data
Coverage1970-2023, 149 systemic banking crises
ReferenceIMF Economic Review, FSB 2023, BIS BCBS 2023

Script Categories

A: Time Series Analysis (6 scripts)
  • A01: All 19 countries grid
  • A02: G7 major economies
  • A03: Credit gap heatmap
  • A04: HP filter decomposition
  • A05: Gap distribution by decade
  • A08: Volatility ranking
B: Signal Extraction (7 scripts)
  • B01: ROC curve
  • B02: Threshold tradeoff
  • B03: Noise-to-signal ratio
  • B04: Precision-recall curve
  • B05: Confusion matrix
  • B06: Lead time sensitivity
  • B07: Crisis timing analysis
C: Machine Learning (8 scripts)
  • C01: AUC comparison
  • C02: ROC overlay
  • C03: Metrics comparison
  • C05-C08: Feature importance (4 models)
D: Panel Analysis (2 scripts)
  • D01: Fixed effects
  • D02: Within vs between variation
E: Statistics (3 scripts)
  • E01: Descriptive statistics
  • E04: Crisis frequency
  • E06: Key statistics infographic

Full Script Inventory

For detailed input/calculation/output documentation with embedded charts, see the full methodology page.