{"id":"W2107105626","doi":"","title":"Optimal unbiased estimators for evaluating agent performance","year":2006,"lang":"en","type":"article","venue":"","topic":"Advanced Bandit Algorithms Research","field":"Decision Sciences","cited_by":23,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Alberta","funders":"","keywords":"Estimator; Variance (accounting); Outcome (game theory); Computer science; Minimum-variance unbiased estimator; Set (abstract data type); Bias of an estimator; Mathematical optimization; Mathematics; Statistics; Mathematical economics; Economics","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":[0.01393105,0.001702849,0.002271584,0.00316637,0.0007281149,0.003039064,0.001436617,0.002312626,0.002416324],"category_scores_gemma":[0.09291584,0.0009069773,0.0007367054,0.001689999,0.002181459,0.004436279,0.002063912,0.001879804,0.0009601251],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001845989,"about_ca_system_score_gemma":0.001925537,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.002226168,"about_ca_topic_score_gemma":0.001568397,"domain_scores_codex":[0.9881363,0.00715071,0.0005681132,0.001202427,0.002317342,0.0006250298],"domain_scores_gemma":[0.9662377,0.02582102,0.002243269,0.002454231,0.00284693,0.0003968868],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"simulation_or_modeling","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.000326989,0.0001014076,0.004327588,0.0002026263,0.0002547987,0.00008448604,0.0001970352,0.7573571,0.001955162,0.1403241,0.002035306,0.09283353],"study_design_scores_gemma":[0.0000273086,0.00007546672,0.0006389005,0.00006763768,0.0000313975,0.00003490769,0.00003734132,0.924674,0.00175585,0.07172813,0.0008998225,0.00002927367],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.01116747,0.0005874933,0.9857931,0.0001963142,0.00003816041,0.00006312709,0.00006199974,0.0002676567,0.001824595],"genre_scores_gemma":[0.5760643,0.00113293,0.4178506,0.0003320472,0.0002149672,0.0006385103,0.0004458675,0.0002789847,0.003041862],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.01393105,"threshold_uncertainty_score":0.07367533,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.2599298327605407,"score_gpt":0.5149763499626402,"score_spread":0.2550465172020995,"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."}}