Supplementary Data S2
=====================

Complete machine-readable outputs supporting the manuscript.

fold_metrics.csv - fold-level predictive metrics
oof_predictions.csv.gz - out-of-fold epoch-level predictions
subject_metrics.csv - subject-level predictive metrics
subject_stage_hyperedge_weights.csv - subject-by-stage temporal hyperedge weights
subject_stage_topology.csv - subject-by-stage topology summaries
hyperedge_stage_tests.csv - Friedman tests with BH-FDR correction
topology_stage_tests.csv - omnibus topology Friedman tests
holm_adjusted_topology_posthoc_tests.csv - pairwise Wilcoxon tests with Holm adjustment
randomized_incidence_null_draws.csv - 1,000 size-preserving randomized-incidence Kendall-W draws, seed 20260817
randomized_incidence_audit_summary.csv - observed/null summaries and empirical p-values
paired_subject_statistics.csv - predictive contrast CIs and Wilcoxon tests
stage_f1.csv - stage-specific out-of-fold F1
MANUSCRIPT_RESULTS.json - machine-readable manuscript result summary
Table1_main_performance.csv - main performance table source

Subjects, not epochs, are the inferential unit for paired statistical comparisons.
