{"id":"W2050257625","doi":"10.1016/j.jclinepi.2007.10.023","title":"Reliability studies of diagnostic tests are not using enough observers for robust estimation of interobserver agreement: a simulation study","year":2008,"lang":"en","type":"article","venue":"Journal of Clinical Epidemiology","topic":"Reliability and Agreement in Measurement","field":"Decision Sciences","cited_by":14,"is_retracted":false,"has_abstract":false,"ca_institutions":"Centre for Advancing Health Outcomes; University of British Columbia","funders":"","keywords":"Kappa; Reliability (semiconductor); Statistics; Sample (material); Variance (accounting); Cohen's kappa; Sample size determination; Variation (astronomy); Inter-rater reliability; Computer science; Population; Variable (mathematics); Econometrics; Mathematics; Medicine","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":["metaresearch"],"category_scores_codex":[0.1036516,0.0002071343,0.003279023,0.0001839471,0.0001273006,0.000006716839,0.0006826267,0.0002001976,0.00003808653],"category_scores_gemma":[0.7784619,0.0001327281,0.0009856916,0.0003570678,0.0006479548,0.0003874422,0.0001635542,0.0003870304,0.000003126933],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0001098127,"about_ca_system_score_gemma":0.0001797572,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00003394821,"about_ca_topic_score_gemma":0.00003536812,"domain_scores_codex":[0.9774507,0.007460346,0.01292408,0.0005308105,0.001335783,0.000298268],"domain_scores_gemma":[0.5078784,0.4641024,0.01991595,0.001119656,0.006739813,0.0002438418],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"simulation_or_modeling","study_design_gemma":"observational","study_design_scores_codex":[0.0004613498,0.001025176,0.4723378,0.00004883952,0.0001266253,0.000002591516,0.0003048152,0.5232999,0.00001243299,0.0000212148,0.0002741004,0.002085094],"study_design_scores_gemma":[0.002309879,0.003254894,0.8557932,0.0002597428,0.0001781822,0.000004112115,0.00114341,0.1263316,0.00003798939,0.01050812,0.00007664211,0.0001021414],"study_design_candidate":"observational","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.940497,0.0002841422,0.05592685,0.001284826,0.001270666,0.0007174982,0.00001162974,0.000003752052,0.000003577909],"genre_scores_gemma":[0.9608097,0.0001804413,0.03854461,0.0002508945,0.0001869694,0.000006316715,8.50485e-7,0.000006782358,0.00001344868],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.6748104,"threshold_uncertainty_score":0.9229794,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.8782298280510391,"score_gpt":0.620131203012851,"score_spread":0.2580986250381881,"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."}}