Rhyo US 12-City Crime Prediction Fairness Audit — Empirical Paper Data Bundle v1.0

Canonical analytical artifacts supporting Surapaneni (2026d), "A 12-City US Cohort Fairness Audit with Cross-City Transfer Evaluation and UK Cross-Program Comparison." Contents include per-city event parquets, modeling-ready feature parquets, all Phase 5.1 (Boston smoke test) and Phase 5.2 (full cohort baseline modeling) outputs, Phase 6 (within-city three-level fairness audit) outputs, Phase 7 (cross-city transfer evaluation) outputs, strengthening-experiment outputs covering R2 threshold sensitivity, R3 architecture invariance, R4 LVMPD external validation, MC multiple-comparisons BH-FDR, SB spatial block bootstrap, E1 feature importance, E3 architecture comparison, NULL test, and UK cross-program comparison, the 16 publication figures in PNG and PDF formats, and the V8 verification pass log. The bundle contains 581 inner files totaling approximately 316 MB compressed. The companion code repository at git tag paper-submission-2026 (commit daa9f5a) is at https://github.com/venkatsurepa/us-city-wide-crime-prediction. The companion OSF retrospective protocol (DOI to be minted) documents the 22-decision chronological tree.本数据集为支撑Surapaneni (2026d)发表的论文《面向美国12个城市的队列公平性审计：跨城市迁移评估与英国跨项目比较》的标准分析制品。内容涵盖各城市事件Parquet（Parquet）文件、可直接用于建模的特征Parquet（Parquet）文件，包含第5.1阶段（波士顿冒烟测试）与第5.2阶段（全队列基准建模）的全部输出、第6阶段（城市内三级公平性审计）输出、第7阶段（跨城市迁移评估）输出，以及覆盖R2阈值敏感性、R3架构不变性、R4拉斯维加斯大都会警察局（LVMPD）外部验证、MC多重比较BH-FDR校正、SB空间块Bootstrap（Bootstrap）、E1特征重要性、E3架构对比、NULL测试与英国跨项目比较的强化实验输出；同时包含16份PNG与PDF格式的出版用图表，以及V8验证通过日志。该数据集打包后共包含581个内部文件，压缩后总大小约316 MB。配套代码仓库的Git标签为paper-submission-2026（提交哈希值：daa9f5a），仓库地址为https://github.com/venkatsurepa/us-city-wide-crime-prediction。配套的开放科学框架（Open Science Framework, OSF）回溯性方案（待分配DOI）记录了包含22个决策节点的时序树。