Aggregate figure source data - second-round revision

Purpose
This archive supplies aggregate, non-identifiable source data for the figures
and validation summaries in the revised manuscript. It contains no patient-level
observations or identifiers.

Figure-source files
- Figure_1_flow_source.csv
- Figure_2A_marker_OR_source.csv
- Figure_2B_corrected_AUC_source.csv
- Figure_2C_delta_AUC_source.csv
- Figure_S1_candidate_audit_source.csv
- Figure_S2_MI_trace_source.csv
- Figure_S3_calibration_source_data.csv
- Figure_S4_score_mortality_source.csv

Aggregate validation summaries
- primary_validation_before_imputation_performance.csv
- primary_validation_before_imputation_differences.csv
- full_cohort_validation_before_imputation_performance.csv
- full_cohort_validation_before_imputation_differences.csv

Direct figure regeneration from aggregate data
1. Install R and the ggplot2 package.
2. Extract the archive without changing its internal file names.
3. Open a terminal in the extracted archive directory and run:
     Rscript 00_regenerate_submitted_figures.R
   This launcher sets locale-safe input and output paths before running the two
   individual plotting scripts.
4. The scripts write PDF, PNG, and TIFF files to the generated_figures folder.

The extracted archive directory is the default input directory. FIGURE_SOURCE_DIR
and FIGURE_OUTPUT_DIR environment variables may be used to override the input
and output directories. The launcher uses relative paths by default, avoiding
locale-dependent handling of non-ASCII absolute paths on Windows.

Derivation of corrected calibration coordinates
The separate statistical-code archive contains the complete ordinary bootstrap
optimism-correction analysis with multiple imputation nested within resampling.
In each iteration, separate MICE runs complete the incomplete bootstrap sample
and the original incomplete cohort used for evaluation. For each model and
performance metric, bootstrap-sample and original-cohort performance are
averaged separately across five imputation indices, and resample-specific
optimism is calculated as the difference between those two averages. The script
derives the corrected calibration curve and confidence-band coordinates and writes
Figure_S3_calibration_source_data.csv. Patient-level data are not redistributed
because they contain protected clinical information.
The reported performance estimates use ordinary bootstrap optimism correction;
they are not out-of-bag, .632, or .632+ estimates.
The present archive is sufficient to reproduce the submitted Figure 2 and
Supplementary Figure S3 from the released aggregate coordinates.

Integrity files
- DATA_DICTIONARY.csv defines all released fields.
- SHA256SUMS.txt records file checksums.
- VALIDATION_REPORT.txt records automated package checks.
