Reproducibility package
Jev router study: scripts and data
Everything behind Testing Jev in High Volume Retrieval and Routing: claim-level data, results, scripts and pre-registrations. The package does not include the dossiers, claim text or evidence text.
Read me first
What each file holds and how the package was built.
| README.md | 8 KB |
Claim-level data
One CSV per test set. Vendor names are replaced by codes (V001 to V153).
| data/devset_calls.csv | 99 KB |
| data/devset_claims.csv | 87 KB |
| data/founder_f3.csv | 4 KB |
| data/pilot.csv | 42 KB |
| data/scaleup.csv | 174 KB |
| data/ws4_planted.csv | 385 KB |
| data/ws4_plants.csv | 3 KB |
Aggregate results
Results files and rendered charts for each workstream.
Scripts
Analysis, scoring and chart scripts. Run python scripts/recount.py from the package folder to recompute the headline counts.
Pre-registrations
Each workstream's pre-registration or protocol, with HASHES.md to check they were not changed after the runs.