{"id":"W4415465292","doi":"10.6000/1929-6029.2025.14.61","title":"The Table 2 Fallacy and Overfitting: A Persistent Problem in Contemporary Research?","year":2025,"lang":"en","type":"article","venue":"International Journal of Statistics in Medical Research","topic":"Health Systems, Economic Evaluations, Quality of Life","field":"Economics, Econometrics and Finance","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"","keywords":"Fallacy; Spurious relationship; Consumption (sociology); Statistical model; Directed acyclic graph; Internal validity; Selection (genetic algorithm); Model selection; Statistical hypothesis testing","routes":{"ca_aff":false,"ca_fund":false,"ca_venue":true,"about_ca":false,"invisible_to_affiliation_only":true},"retraction":null,"screen":null,"direct_labels":[],"prediction":{"model_version":"metacan-v3-hybrid-931329e0061c","candidate_categories":["metaresearch"],"consensus_categories":["metaresearch"],"category_scores_codex":[0.4581949,0.002190995,0.006802405,0.007106764,0.003122102,0.008551018,0.007913739,0.007265997,0.009913805],"category_scores_gemma":[0.7332757,0.002577796,0.004829697,0.01078034,0.02108677,0.01263744,0.005816939,0.01390901,0.002298566],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.004714729,"about_ca_system_score_gemma":0.008358991,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.007806568,"about_ca_topic_score_gemma":0.007985158,"domain_scores_codex":[0.5733888,0.3292822,0.02842764,0.03020977,0.03733824,0.001353418],"domain_scores_gemma":[0.1576701,0.7633251,0.01729345,0.04602239,0.01463058,0.001058387],"domain_codex":null,"domain_gemma":"methods","domain_candidate":"methods","domain_consensus":null,"study_design_codex":"design_other","study_design_gemma":"theoretical_or_conceptual","study_design_scores_codex":[0.00126922,0.0002251411,0.03371124,0.008276048,0.004732871,0.001198485,0.01071339,0.008648957,0.0005491627,0.3449188,0.1167277,0.4690289],"study_design_scores_gemma":[0.0004494113,0.000302591,0.005563742,0.004940433,0.000648991,0.000859028,0.001551903,0.01808753,0.0006560058,0.8992893,0.06738646,0.0002646078],"study_design_candidate":"theoretical_or_conceptual","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"commentary","genre_scores_codex":[0.008928444,0.01903432,0.7869661,0.1590178,0.01244457,0.001728826,0.00209486,0.001623435,0.008161703],"genre_scores_gemma":[0.1890377,0.01105366,0.6266611,0.1471523,0.01041759,0.008634595,0.001802298,0.00122922,0.00401161],"genre_candidate":"commentary","genre_consensus":null,"teacher_disagreement_score":0.5418051,"threshold_uncertainty_score":0.6681422,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.559293903084334,"score_gpt":0.5948790970291169,"score_spread":0.03558519394478288,"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."}}