GLP-1 Virtual Cohort v1

This upload contains the full technical documentation and a free evaluation sample (498 patients, 6 branches, 104 weeks each — 310,752 weekly records) — not the full dataset. The complete dataset — 20,000 patients, 12,480,000 weekly records, 104 weeks of weekly follow-up across 6 scenario branches — is available at: https://sentineldata.com.ua/dataset/glp-1-virtual-cohort Sample contents: sample_patients_v1.parquet 498 rows sample_weekly_as_treated_v1.parquet 51,792 rows sample_weekly_switch_drug_v1.parquet 51,792 rows sample_weekly_dose_variation_v1.parquet 51,792 rows sample_weekly_early_discontinuation_v1.parquet 51,792 rows sample_weekly_treatment_holiday_v1.parquet 51,792 rows sample_weekly_no_treatment_v1.parquet 51,792 rows CSV mirrors are not included in this sample; Parquet is the canonical format. A short guide on opening Parquet files ( HOW_TO_OPEN_PARQUET.txt ) is included. The sample is a real subset of the full dataset — drawn by stratified (per-molecule) random sampling with seed 42, exactly 83 patients per molecule (6 × 83 = 498), with complete weekly records in all 6 scenario branches for every sampled patient. This is an extract, not a re-generation: every row was identity-checked against the full dataset. Unlike a naive first-N-by-ID slice, this sample is balanced across all 6 molecules, so it is representative of the full cohort's molecule mix. Fully ... Full description with all tables is on the Wiki page of this project. Full dataset and purchase: https://sentineldata.com.ua/dataset/glp-1-virtual-cohort . Zenodo copy: https://doi.org/10.5281/zenodo.22998938