Stroke Time-to-Treatment Counterfactual (200,000 Patients)

This upload contains the full technical documentation and a free evaluation sample (1,000 patients, 1,424 branch rows) — not the full dataset. The complete dataset — 200,000 patients, 302,400 branch rows, 33 columns, with the full generator code — is available at: https://sentineldata.com.ua/dataset/stroke-time-to-treatment-counterfactual Sample contents: Item Value File stroke_counterfactual_sample.parquet (inside the sample archive, with the sample passport in Russian) Patients 1,000 (STK-000001 to STK-001000) Branch rows 1,424: 1,000 baseline + 284 tPA (71 patients × 4) + 140 thrombectomy (35 patients × 4) Columns 33, same schema as the full table, including the 7 latent ground-truth columns Stroke type 798 ischemic / 202 hemorrhagic Nulls 0 The sample is a true subset: the first 1,000 patients by ID with all of their branch rows, not a stratified or randomized draw, produced by the same generation code as the full dataset. Because it is only 1,000 patients, it contains fewer treated patients than the full dataset’s proportions (7.1% tPA and 3.5% thrombectomy, versus 8.4% and 4.4% in the full 200,000). The generator code and resolved parameters are part of the full product and are not included here. Documentation in this upload: the full data passport and the sample passport (PDF, English), schema_stroke_counterfactual.md (data dictionary), validation_report.md (full ... Full description with all tables is on the Wiki page of this project. Full dataset and purchase: https://sentineldata.com.ua/dataset/stroke-time-to-treatment-counterfactual . Zenodo copy: https://doi.org/10.5281/zenodo.23088287