{"id":"W2768609493","doi":"10.1097/ede.0000000000000787","title":"Can We Train Machine Learning Methods to Outperform the High-dimensional Propensity Score Algorithm?","year":2017,"lang":"en","type":"article","venue":"Epidemiology","topic":"Statistical Methods and Inference","field":"Mathematics","cited_by":64,"is_retracted":false,"has_abstract":true,"ca_institutions":"McGill University Health Centre; McGill University; University of British Columbia; Centre for Advancing Health Outcomes; Jewish General Hospital; Providence Health Care","funders":"Canadian Institutes of Health Research","keywords":"Propensity score matching; Confounding; Machine learning; Covariate; Computer science; Selection bias; Artificial intelligence; Random forest; Algorithm; Elastic net regularization; Feature selection; Statistics; Mathematics","routes":{"ca_aff":true,"ca_fund":true,"ca_venue":false,"about_ca":false,"invisible_to_affiliation_only":false},"retraction":null,"screen":null,"direct_labels":[],"prediction":{"model_version":"codex-gemma-dda1882f352a","candidate_categories":["metaresearch"],"consensus_categories":[],"category_scores_codex":[0.0106842,0.000220699,0.0007989949,0.00003912642,0.0008535207,0.0000249476,0.0005611154,0.0001504053,0.0003671273],"category_scores_gemma":[0.089315,0.0001253121,0.00009449169,0.00004819466,0.0003457383,0.00003643964,0.0003874428,0.0006974015,0.00004582852],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00004743964,"about_ca_system_score_gemma":0.00006283793,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.001109505,"about_ca_topic_score_gemma":0.0001740181,"domain_scores_codex":[0.9955956,0.002807096,0.0005442005,0.0004212485,0.0001193484,0.0005124749],"domain_scores_gemma":[0.9843526,0.01423788,0.0003523671,0.0007545447,0.000105365,0.0001972228],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"design_other","study_design_gemma":"theoretical_or_conceptual","study_design_scores_codex":[0.00002737158,0.00003097589,0.00529401,0.00002229316,0.00004527144,0.000009454141,0.0003189746,0.00002596736,0.000214859,0.3172935,0.002089487,0.6746279],"study_design_scores_gemma":[0.0002399918,0.0002463875,0.04696065,0.00004860778,0.00004147534,0.00003242008,0.00002385992,0.03059109,0.000303361,0.916869,0.004425265,0.0002178356],"study_design_candidate":"theoretical_or_conceptual","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"methods","genre_scores_codex":[0.03331717,0.0001175615,0.9161416,0.04798517,0.0007143816,0.0004776795,0.0000551446,0.00007129331,0.001119969],"genre_scores_gemma":[0.04433944,0.00001662945,0.9519727,0.002262837,0.0002508301,0.00004441475,0.000005942551,0.00002360681,0.00108358],"genre_candidate":"methods","genre_consensus":"methods","teacher_disagreement_score":0.67441,"threshold_uncertainty_score":0.9183561,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.3766227571966274,"score_gpt":0.4804717655727531,"score_spread":0.1038490083761257,"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."}}