{"id":"W6930214893","doi":"10.5281/zenodo.11267987","title":"Data from: Understanding Test Convention Consistency as a Dimension of Test Quality","year":2024,"lang":"en","type":"dataset","venue":"Zenodo (CERN European Organization for Nuclear Research)","topic":"Fecal contamination and water quality","field":"Environmental Science","cited_by":1,"is_retracted":false,"has_abstract":true,"ca_institutions":"McGill University","funders":"","keywords":"Consistency (knowledge bases); Test (biology); Dimension (graph theory); Convention; Quality (philosophy); Test data; Software","routes":{"ca_aff":true,"ca_fund":false,"ca_venue":false,"about_ca":false,"invisible_to_affiliation_only":false},"retraction":null,"screen":null,"direct_labels":[],"prediction":{"model_version":"metacan-v3-hybrid-931329e0061c","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.002543709,0.002341937,0.00141558,0.004205052,0.0009572775,0.0031988,0.003462101,0.003093092,0.0396039],"category_scores_gemma":[0.01823165,0.0007779851,0.001651615,0.006513205,0.0006049198,0.001758315,0.00306243,0.002671006,0.05606974],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.002408363,"about_ca_system_score_gemma":0.002803335,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.03623127,"about_ca_topic_score_gemma":0.06272609,"domain_scores_codex":[0.9971723,0.0006170188,0.0004079443,0.0006217776,0.0008480911,0.0003329326],"domain_scores_gemma":[0.9908601,0.003498012,0.001014679,0.00214824,0.001914252,0.0005646705],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"not_applicable","study_design_gemma":"not_applicable","study_design_scores_codex":[0.00008403767,0.0000317953,0.001964216,0.000784626,0.0000399443,0.00002433512,0.00003486819,0.0006457617,0.0001008154,0.0007648879,0.9923623,0.003162414],"study_design_scores_gemma":[0.0003826035,0.00003135109,0.01496762,0.0007134456,0.00006193791,0.0001013344,0.0001434768,0.001477633,0.0005432991,0.004579355,0.9769337,0.00006427118],"study_design_candidate":"not_applicable","study_design_consensus":"not_applicable","genre_codex":"dataset","genre_gemma":"dataset","genre_scores_codex":[0.0001804116,0.00007790895,0.0001255909,0.0001127422,0.00002523943,0.000009452068,0.9986495,0.0002802523,0.0005388918],"genre_scores_gemma":[0.0005019626,0.00004111457,0.0003317637,0.00004882761,0.000005582191,0.00005569707,0.9986743,0.00004879297,0.000291966],"genre_candidate":"dataset","genre_consensus":"dataset","teacher_disagreement_score":0.0396039,"threshold_uncertainty_score":0.1324882,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.1576281945584947,"score_gpt":0.3249054793219208,"score_spread":0.1672772847634262,"validation_status":"score_only:v0-immature-baseline","note":"Baseline scores from an immature model (maturity gate not passed). Scores rank; they never assert a category."}}