```json { "table_name": "ae", "primary_key": "USUBJID,AESEQ", "candidates": [ { "rank": 0, "partition_key": "USUBJID", "partition_key_label": "E", "partition_key_reasoning": "USUBJID is the entity column (subject/patient ID) with RPUV=5.29 (225 unique subjects across 1,191 adverse events), Top 3 Coverage=4.87%, making it an excellent partition key. Each subject has multiple adverse events (mean ~5.3 events per subject), which is ideal for grouping related events together.", "clustering_key": "AEDTC", "clustering_key_label": "S", "clustering_key_reasoning": "AEDTC is the adverse event date/time with monotonicity=0.489 and 453 unique values. This temporal column orders events chronologically within each subject's timeline. AESEQ (sequence number) could also work, but AEDTC provides actual temporal ordering which is more meaningful for adverse event analysis.", "world_line": "USUBJID", "session_column": null, "confidence": "HIGH", "notes": "This is a clinical trial adverse events table (CDISC SDTM AE domain). Partitioning by USUBJID groups all adverse events per subject, and clustering by AEDTC orders them chronologically. The composite primary key (USUBJID, AESEQ) uniquely identifies each adverse event record. AESEQ could be used for clustering instead of AEDTC if strict sequence ordering is preferred over temporal ordering." } ] } ```