Document index
This index lists the public, reader-facing summaries of the locked higher-ed evidence. Each summary is a short, plain-language version of the underlying internal document; no per-institution numbers and no raw data are linked from these summaries.
Lock and claim discipline
Final Lock Note
What the higher-ed extension is, what it is not, the validated lane, the headline result, and the Phase-19 hardening evidence. Single-page reference.
Read the Final Lock Note →Claim Boundary Note
Allowed and forbidden claims. The 13-item boundary checklist that must accompany any external restatement of the result.
Read the Claim Boundary Note →Phase 19 Hardening Decision
Integrated decision across the five Phase-19 hardening tracks; final classification and recommended next focus.
Read the Phase 19 Decision →Validation tracks
Cluster Bootstrap
Institution-level (UNITID) bootstrap of the closure AUROC; cluster CI vs analytic CI; paired-difference distribution.
Read the Cluster Bootstrap Note →Temporal Folds
Four sliding train/test time-block folds; joint-weak AUROC across folds; honest record of the Fold-1 baseline tie.
Read the Temporal Folds Note →Closure Lead-Time
Score distributions and top-decile capture rates across T-1 through T-5 in the validated lane.
Read the Lead-Time Analysis →ED Financial Responsibility Baseline
Federal composite score as a comparator on the descriptive train cohort; the head-to-head test on the locked cohort is not runnable due to a data-coverage gap.
Read the ED FR Baseline Note →Heightened Cash Monitoring Anchor
Source-blocked in this run; documented as a candidate for future work.
Read the HCM Anchor Note →Reproducibility
The locked higher-ed result is reproducible from documented Python scripts and public IPEDS data. The full reproducibility runbook, the artifact index marking files as CURRENT / HISTORICAL / SUPERSEDED, and the curated minimum-evidence packet are part of the project repository. They are not republished here as standalone files because they describe internal paths; the same information is summarised in the document pages above.
The project records SHA-256 fingerprints for every raw IPEDS zip (per the per-phase fetch logs) and an MD5 lock baseline for the locked-evidence documents and CSV outputs. A reviewer who needs to verify the lock can compute the same fingerprints from the public IPEDS sources.