{"id":"W2158460222","doi":"10.1177/0272989x09341755","title":"The Relative Ability of Different Propensity Score Methods to Balance Measured Covariates Between Treated and Untreated Subjects in Observational Studies","year":2009,"lang":"en","type":"article","venue":"Medical Decision Making","topic":"Advanced Causal Inference Techniques","field":"Mathematics","cited_by":456,"is_retracted":false,"has_abstract":true,"ca_institutions":"Institute for Clinical Evaluative Sciences; University of Toronto","funders":"","keywords":"Propensity score matching; Covariate; Statistics; Observational study; Inverse probability weighting; Weighting; Medicine; Mathematics","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":"codex-gemma-dda1882f352a","candidate_categories":["metaresearch"],"consensus_categories":[],"category_scores_codex":[0.003171956,0.000192428,0.0006584102,0.00008431691,0.0001423613,0.00001455837,0.000233614,0.000155353,0.00001094927],"category_scores_gemma":[0.06082897,0.0001089966,0.0000436767,0.0004450858,0.0001962963,0.0001081311,0.0001788584,0.0003462847,5.832123e-7],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0001639239,"about_ca_system_score_gemma":0.00004930103,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.000007983872,"about_ca_topic_score_gemma":0.0000727966,"domain_scores_codex":[0.9972337,0.0007146614,0.0007686635,0.000324212,0.0007173566,0.0002414251],"domain_scores_gemma":[0.9799435,0.01901354,0.0002406182,0.0003129677,0.0003964712,0.00009297119],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"observational","study_design_gemma":"theoretical_or_conceptual","study_design_scores_codex":[0.000550431,0.0002041216,0.5170233,0.00009579607,0.000130781,0.00002638149,0.002262875,0.000004978218,0.004319331,0.07273739,0.0001806051,0.4024641],"study_design_scores_gemma":[0.0002366698,0.0001218855,0.4817345,0.0007924183,0.00001784219,7.830027e-7,0.00003355814,0.0001983079,0.002242804,0.5145416,0.000006351394,0.00007327541],"study_design_candidate":"observational","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.7554493,0.0001682441,0.2431628,0.0005098166,0.00002829184,0.000562806,0.000002969732,0.00006567833,0.00005006706],"genre_scores_gemma":[0.8561302,0.00006013981,0.1436714,0.00007800737,0.00001832942,0.00002335976,0.000001626554,0.000008071373,0.000008924541],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.4418042,"threshold_uncertainty_score":0.947082,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.4399813510196111,"score_gpt":0.5151726562683904,"score_spread":0.07519130524877926,"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."}}