PASS hash E107_GATES.json PASS hash E107_REPORT.md PASS hash PREREGISTRATION.json PASS hash rows.json PASS hash summary.json PASS hash verify_e107_offline.py PASS edge_lock: same tasks in both arms PASS edge_lock: success recount PASS edge_lock: McNemar recomputed PASS edge_lock: step sign test recomputed PASS grid_key_door: same tasks in both arms PASS grid_key_door: success recount PASS grid_key_door: McNemar recomputed PASS grid_key_door: step sign test recomputed PASS grid_other: same tasks in both arms PASS grid_other: success recount PASS grid_other: McNemar recomputed PASS grid_other: step sign test recomputed PASS practice edge: counts PASS practice edge: used rows have exactly one candidate and a complete census PASS practice edge: rows whose cause lies outside its own support = 1, all disclosed PASS practice edge: gate re-derived without those rows (feature ['col', 0], p 0.001, usable True) matches the disclosed sensitivity PASS practice edge: oracle accuracy recount PASS practice edge: gate re-derived independently (z 14.472, feature ['col', 0], p 0.001, usable True) PASS practice grid_kd: counts PASS practice grid_kd: used rows have exactly one candidate and a complete census PASS practice grid_kd: rows whose cause lies outside its own support = 0, all disclosed PASS practice grid_kd: oracle accuracy recount PASS practice grid_kd: gate re-derived independently (z 3.803, feature ['col', 0], p 0.018, usable False) PASS H1 verdict PASS H1b verdict PASS H2 verdict PASS H3 verdict PASS H4 verdict PASS H5 verdict 35/35 checks passed