{"id":"W4318067345","doi":"10.1002/sim.9665","title":"Measuring the performance of prediction models to personalize treatment choice","year":2023,"lang":"en","type":"article","venue":"Statistics in Medicine","topic":"Advanced Causal Inference Techniques","field":"Mathematics","cited_by":43,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"Medical Research Council; Medical Research Council Canada; ZonMw; European Commission; Schweizerischer Nationalfonds zur Förderung der Wissenschaftlichen Forschung; National Science Foundation","keywords":"Computer science; Covariate; Calibration; Machine learning; Baseline (sea); Outcome (game theory); Range (aeronautics); Artificial intelligence; Predictive modelling; Data mining; Statistics; Mathematics","routes":{"ca_aff":false,"ca_fund":true,"ca_venue":false,"about_ca":false,"invisible_to_affiliation_only":true},"retraction":null,"screen":null,"direct_labels":[{"model":"gpt","categories":[],"domain":null,"study_design":"simulation_or_modeling","genre":"methods","about_ca_system":false,"about_ca_topic":false,"confidence":"high","status":"direct model label, unvalidated"},{"model":"grok","categories":[],"domain":null,"study_design":"simulation_or_modeling","genre":"methods","about_ca_system":false,"about_ca_topic":false,"confidence":"high","status":"direct model label, unvalidated"},{"model":"opus","categories":[],"domain":null,"study_design":"simulation_or_modeling","genre":"methods","about_ca_system":false,"about_ca_topic":false,"confidence":"high","status":"direct model label, unvalidated"}],"prediction":{"model_version":"codex-gemma-dda1882f352a","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.0005356165,0.0001138824,0.0002339715,0.0001383375,0.00004685373,0.000002604024,0.0001325531,0.00003263298,0.00003490682],"category_scores_gemma":[0.0007846156,0.00007208513,0.00001234872,0.0003874168,0.00009801732,0.00005173915,0.00003392439,0.000101718,0.000005889469],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0001456818,"about_ca_system_score_gemma":0.00002871421,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0001296603,"about_ca_topic_score_gemma":0.0001136607,"domain_scores_codex":[0.9989575,0.00004243644,0.0003269897,0.0001415687,0.0003476017,0.0001839741],"domain_scores_gemma":[0.9984701,0.001043003,0.00008557617,0.0002554962,0.0001022855,0.00004359917],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"theoretical_or_conceptual","study_design_gemma":"theoretical_or_conceptual","study_design_scores_codex":[0.000210655,0.0003119239,0.01195835,0.001149654,0.0001745661,0.00005054548,0.05671098,0.01365816,0.01231369,0.766417,0.04339752,0.09364694],"study_design_scores_gemma":[0.001377059,0.002261954,0.009245682,0.001252796,0.000121362,0.00000882517,0.002547754,0.2169482,0.006460429,0.7580605,0.001439485,0.000275905],"study_design_candidate":"theoretical_or_conceptual","study_design_consensus":"theoretical_or_conceptual","genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.4817916,0.0001207281,0.5119709,0.0009334237,0.0002778482,0.001176931,0.0003156766,0.0003825582,0.003030322],"genre_scores_gemma":[0.929494,0.00049017,0.06905901,0.00006396721,0.0001127807,0.000157927,0.00002630869,0.00002561219,0.0005702826],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.4477023,"threshold_uncertainty_score":0.2939546,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.2535139144086483,"score_gpt":0.4251422307621626,"score_spread":0.1716283163535143,"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."}}