# Run 2 Temporal Analysis (Hypothesis-Conditioned) This report is intentionally **hypothesis-conditioned**: it models temporal signatures under the assumption that non-focal groups could include synthetic/back-dated filings. It does **not** conclude that this is true; it surfaces patterns that would be notable if the hypothesis were true. ## Dataset Scope - Unique docs: **1,551** - Tracking-font object rows: **3,743** - Clusters: **78** - Cases: **216** - Tracking hashes: **12** - Tracking groups: **12** ## Focal Anchor - Focal cluster ID: **1570** - Focal docs: **45** - Focal filing window: **2023-01-24 → 2025-07-08** - Non-focal docs: **1,506** - Non-focal docs dated earlier than focal start (2023-01-24): **682** ## Top Temporal Signals by Group Columns: docs, focal share, first non-focal vs first focal gap, pre-focal non-focal %, span days, composite score. ``` object_pdf_group n_docs_all n_docs_focal n_docs_nonfocal first_nonfocal first_focal gap_nonfocal_before_focal_days pre_entity_nonfocal_pct span_all_days temporal_pressure_score B1 254 27 227 2017-02-21 2023-06-14 2304 76.211454 3104 92.500000 A1 65 1 64 2017-01-19 2023-01-24 2196 56.250000 2490 71.666667 C1 1130 31 1099 2020-01-07 2023-07-13 1283 57.324841 2003 66.666667 C2 1130 31 1099 2020-01-07 2023-07-13 1283 57.324841 2003 66.666667 A2 133 5 128 2020-01-06 2023-01-24 1114 75.781250 1970 63.750000 A4 17 5 12 2020-02-05 2023-01-24 1084 83.333333 1940 62.500000 A3 134 5 129 2020-01-06 2023-01-24 1114 75.193798 1970 58.750000 D2 500 20 480 2021-08-13 2023-11-15 824 82.708333 1354 54.166667 ``` ## Top Temporal Signals by Hash ``` object_pdf_group object_sha256 object_description n_docs_all n_docs_focal n_docs_nonfocal gap_nonfocal_before_focal_days pre_entity_nonfocal_pct span_all_days temporal_pressure_score B1 427af119e8f704848644710f5da65ea8d385162ce3d78bb774a2b0bda4893f95 TRACKING-FONT-B1 254 27 227 2304 76.211454 3104 92.500000 A1 e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855 TRACKING-FONT-A1 65 1 64 2196 56.250000 2490 71.666667 C1 1fa67c75ff9672377725313c162f83635e9ba546d7c5b3e218008aa88edecef8 TRACKING-FONT-C1 1130 31 1099 1283 57.324841 2003 66.666667 C2 b50a0345994f437ffae96470d96a83c2efd26724aa15445c3c7068ad26e17692 TRACKING-FONT-C2 1130 31 1099 1283 57.324841 2003 66.666667 A2 3909d3a28c60f88a605fb4943057672986eb0283f63fe41c64052962f2565e0f TRACKING-FONT-A2 133 5 128 1114 75.781250 1970 63.750000 A4 10f3aaa3db59a93ed277f15438bd1472631ff600f829cac4f1b5fd21f378f4d4 TRACKING-FONT-A4 17 5 12 1084 83.333333 1940 62.500000 A3 f627466c20a82cd207d50833a58a7100909b3f5de2c6df5303ad7c16c4f6dcdc TRACKING-FONT-A3 134 5 129 1114 75.193798 1970 58.750000 D2 086b4c456dbe6f09f8c99517bcda6cd0f82646bc0cf77bf0a1e64dd9a7c38fd3 TRACKING-FONT-D2 500 20 480 824 82.708333 1354 54.166667 ``` ## Longest Cluster Time Ranges ``` cluster_id_num cluster_name n_docs n_hashes n_pdf_groups first_filing last_filing span_days is_focal_cluster 290.0 ADRIAN WESLEY 14 5.0 5.0 2017-01-19 2024-03-19 2616 False 230.0 IFRAH HASSAN 27 9.0 9.0 2018-08-02 2024-04-03 2071 False 80.0 AESHA OSMAN 59 9.0 9.0 2018-07-23 2024-03-20 2067 False 1726.0 RAMADAN CAMPBELL 13 6.0 6.0 2018-07-23 2024-02-09 2027 False 193.0 WILLIAM NABORS 15 7.0 7.0 2018-11-01 2024-01-25 1911 False 14.0 TERRELL JOHNSON 80 8.0 8.0 2019-06-25 2024-03-12 1722 False 147.0 ANGELIC NUNN 66 9.0 9.0 2020-02-05 2024-04-11 1527 False 267.0 REX BASSWOOD 41 10.0 10.0 2020-03-11 2024-04-22 1503 False 493.0 BISHARO NOOR 13 5.0 5.0 2020-04-03 2024-04-11 1469 False 76.0 MAKIS LANE 91 6.0 6.0 2020-04-09 2024-04-05 1457 False ``` ## Yearly Volume (focal vs non-focal) ``` year nonfocal_docs focal_docs total_docs 2017 6 0 6 2018 6 0 6 2019 17 0 17 2020 47 0 47 2021 187 0 187 2022 384 0 384 2023 685 4 689 2024 156 8 164 2025 18 33 51 ``` ## QA / Limits - In this run's tables, `filing_date` and `efile_date` are identical for all docs, so this run cannot independently validate backdating from those two fields alone. - Temporal pressure score is a prioritization heuristic (rank-based), not a proof metric. - Strong signals indicate where to inspect provenance metadata, signature chains, and raw ingestion timestamps next.