Data & Code Downloads

Publication Tables

• 📊
  table1_descriptive_stats.csv

  Table 1: Descriptive statistics for all variables (mean, SD, range, coverage)
• 📊
  table2_evidence_hierarchy.csv

  Table 2: Evidence hierarchy across all analytical methods
• 📊
  table3_bym2_robustness.csv

  Table 3: BYM2 robustness across model specifications (v1, v2A, v2B)
• 📊
  table4_compound_bym2.csv

  Table 4: Compound-specific BYM2 results (12 compounds × 2 cancers)
• 📊
  table5_long_difference.csv

  Table 5: Long-difference results (Δpesticide 1997→2012)
• 📊
  table6_gauntlet_summary.csv

  Table 6: Cross-gauntlet summary (4 risk factors, IARC scores, pesticide survival)

Risk Factor Gauntlet Results

• 📊
  smoking_bym2_results.csv

  Smoking gauntlet: BYM2 results for 12 cancer types
• 📊
  smoking_long_diff_results.csv

  Smoking gauntlet: long-difference results
• 📊
  smoking_iv_results.csv

  Smoking gauntlet: IV/2SLS results
• 📊
  smoking_vs_pesticide_comparison.csv

  Smoking vs pesticide head-to-head comparison
• 📊
  obesity_bym2_results.csv

  Obesity gauntlet: BYM2 results for 12 cancer types
• 📊
  obesity_long_diff_results.csv

  Obesity gauntlet: long-difference results
• 📊
  obesity_iv_results.csv

  Obesity gauntlet: IV/2SLS results
• 📊
  obesity_vs_pesticide_comparison.csv

  Obesity vs pesticide head-to-head comparison
• 📊
  alcohol_bym2_results.csv

  Alcohol gauntlet: BYM2 results for 12 cancer types
• 📊
  alcohol_long_diff_results.csv

  Alcohol gauntlet: long-difference results
• 📊
  alcohol_iv_results.csv

  Alcohol gauntlet: IV/2SLS results
• 📊
  alcohol_vs_pesticide_comparison.csv

  Alcohol vs pesticide head-to-head comparison
• 📊
  inactivity_bym2_results.csv

  Physical inactivity gauntlet: BYM2 results for 12 cancer types
• 📊
  inactivity_long_diff_results.csv

  Inactivity gauntlet: long-difference results
• 📊
  inactivity_iv_results.csv

  Inactivity gauntlet: IV/2SLS results
• 📊
  inactivity_vs_pesticide_comparison.csv

  Inactivity vs pesticide head-to-head comparison

Exploratory Screening

• 📊
  exploratory_synthesis.csv

  Full exploratory synthesis: all predictors × 26 cancers (Spearman, partial, LASSO, OLS)
• 📊
  top_correlates_by_cancer.csv

  Top predictor correlates for each of 26 cancer types

Core Analysis Tables

• 📊
  spearman_correlation_matrix.csv

  Full pairwise Spearman rank correlation matrix across all variables
• 📊
  spearman_pvalue_matrix.csv

  P-value matrix for all pairwise Spearman correlations
• 📊
  top_correlations.csv

  Top predictor–cancer correlations ranked by absolute magnitude
• 📊
  partial_correlations_cancer.csv

  Partial correlations controlling for 9 confounders
• 📊
  vif_scores.csv

  Variance Inflation Factors for multicollinearity assessment
• 📊
  morans_i_results.csv

  Global Moran’s I spatial autocorrelation statistics for each variable
• 📊
  data_coverage_report.csv

  Variable-level coverage report: non-null counts, missing percentages

Bayesian & Compound-Specific Results

• 📊
  all_cancer_type_models.csv

  BYM2 results for all 8 cancer types: rate ratios, HDIs, rho, diagnostics
• 📊
  all_compound_bym2_results.csv

  Compound-specific BYM2: 12 compounds × 2 cancers = 24 models
• 📊
  expanded_compound_bym2_results.csv

  Expanded compound BYM2: 6 additional compounds (3 herbicides + 3 fungicides)
• 📊
  exposure_pathway_results.csv

  Urban/rural stratified BYM2, interaction models, crop sensitivity

Robust Analysis

• 📊
  spatial_models_all_cancers.csv

  OLS, Spatial Lag, and Spatial Error model comparisons across cancer types
• 📊
  negative_control_tests.csv

  Negative control results: lifestyle, random, and unrelated exposure tests

Sensitivity Analysis

• 📊
  bootstrap_correlations.csv

  Bootstrap confidence intervals (1,000 resamples) for Spearman correlations
• 📊
  bootstrap_regression.csv

  Bootstrap confidence intervals for OLS regression coefficients
• 📊
  leave_one_state_out.csv

  Jackknife estimates excluding each state in turn (50 iterations)
• 📊
  lag_window_sensitivity.csv

  Sensitivity to different exposure–outcome temporal lag windows
• 📊
  suppression_sensitivity.csv

  Sensitivity to varying county-inclusion thresholds (NCI suppression)
• 📊
  outlier_cooks_comparison.csv

  Outlier sensitivity: Cook’s distance based trimming
• 📊
  outlier_iqr_comparison.csv

  Outlier sensitivity: IQR-based trimming

Synthesis & Summary

• 📊
  methods_comparison.csv

  All analytical methods compared: effect estimates, spatial controls, interpretations
• 📊
  confounding_robustness.csv

  Confounding robustness: RR stability across 5 model specifications

Reproducibility

The complete analysis pipeline consists of 32 Jupyter notebooks and 5 Python source modules:

• src/data_collection.py — Data download functions (~3,400 lines)
• src/cleaning.py — FIPS standardization, merge logic (~1,300 lines)
• src/analysis.py — All statistical methods (~2,500 lines)
• src/visualization.py — Plotly maps, forest plots, figures (~1,700 lines)
• src/sensitivity.py — Bootstrap, jackknife, robustness tests (~540 lines)