{"id":"W2566903824","doi":"10.1016/j.nicl.2016.12.018","title":"Selection bias in the reported performances of AD classification pipelines","year":2016,"lang":"en","type":"article","venue":"NeuroImage Clinical","topic":"Bayesian Modeling and Causal Inference","field":"Computer Science","cited_by":35,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"FP7 Information and Communication Technologies; Seventh Framework Programme; Engineering and Physical Sciences Research Council; National Institute of Biomedical Imaging and Bioengineering; Canadian Institutes of Health Research; Genentech; National Institutes of Health; Eisai; Servier; U.S. Department of Defense; Eli Lilly and Company; Lundbeckfonden; National Institute on Aging; National Institute for Health and Care Research; Pfizer; BioClinica; Biogen; Novartis Pharmaceuticals Corporation; Bristol-Myers Squibb Foundation; F. Hoffmann-La Roche; Merck; Alzheimer's Drug Discovery Foundation; University College London; IXICO; Takeda Pharmaceutical Company; AbbVie; Fujirebio Europe; Alzheimer's Association; Foundation for the National Institutes of Health; University College London Hospitals NHS Foundation Trust; GE Healthcare; Alzheimer's Disease Neuroimaging Initiative; Medical Research Council; Johnson and Johnson; Meso Scale Diagnostics","keywords":"Resampling; Pipeline (software); Selection bias; Pipeline transport; Computer science; Consistency (knowledge bases); Selection (genetic algorithm); Machine learning; Artificial intelligence; Sampling bias; Fraction (chemistry); Data mining; Statistics; Sample size determination; Engineering; 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":[],"prediction":{"model_version":"codex-gemma-dda1882f352a","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.001737128,0.00007597212,0.0001404312,0.00005882349,0.00003988184,0.00003818618,0.0005469166,0.00006555012,0.000005100489],"category_scores_gemma":[0.001081903,0.0000401524,0.00006670804,0.0003496601,0.0001051603,0.0003068553,0.00004993718,0.0001800524,0.00002275964],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.000006180776,"about_ca_system_score_gemma":0.00007525403,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.000005981304,"about_ca_topic_score_gemma":0.00001114927,"domain_scores_codex":[0.9983002,0.000326732,0.0006870757,0.0003345886,0.0002171387,0.0001342473],"domain_scores_gemma":[0.9984248,0.0006900005,0.0002509909,0.0004753729,0.0001280939,0.00003071067],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"design_other","study_design_gemma":"observational","study_design_scores_codex":[0.00005802539,0.0005012528,0.167888,0.0000198971,0.000009725802,0.00002689425,0.000319352,0.00001843671,0.01738444,0.007092363,0.002648935,0.8040326],"study_design_scores_gemma":[0.0005424966,0.0003978898,0.7943192,0.00006662932,0.000009624016,0.00006204636,0.00002144092,0.1977102,0.001295464,0.003401151,0.002023976,0.0001499948],"study_design_candidate":"observational","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.6471187,0.0000224062,0.3463714,0.005394847,0.0003302922,0.0001001077,7.29148e-7,0.00006932476,0.0005921522],"genre_scores_gemma":[0.9958798,0.0001813704,0.003398505,0.0003793244,0.000068968,0.000007097772,4.789161e-7,0.000003747973,0.00008069412],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.8038827,"threshold_uncertainty_score":0.1637368,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.3111626821699292,"score_gpt":0.4087706249836011,"score_spread":0.09760794281367186,"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."}}