{"id":"W4390929439","doi":"10.1098/rsos.231003","title":"Replicability of simulation studies for the investigation of statistical methods: the RepliSims project","year":2024,"lang":"en","type":"article","venue":"Royal Society Open Science","topic":"Scientific Computing and Data Management","field":"Decision Sciences","cited_by":13,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Toronto; York University","funders":"Leids Universitair Medisch Centrum; ZonMw","keywords":"Computer science; Replication (statistics); Code (set theory); Data science; Visualization; Empirical research; Information retrieval; Data mining; Statistics; Programming language","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":["metaresearch"],"consensus_categories":["metaresearch"],"category_scores_codex":[0.6760364,0.004497476,0.008347963,0.01068146,0.004211219,0.01357596,0.00869853,0.007714557,0.02300391],"category_scores_gemma":[0.9148342,0.003772319,0.01440775,0.01079621,0.01448554,0.01163515,0.01720053,0.01178205,0.003493189],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00772949,"about_ca_system_score_gemma":0.02300652,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.002221418,"about_ca_topic_score_gemma":0.001256495,"domain_scores_codex":[0.2070998,0.7003319,0.03667709,0.01868415,0.03521655,0.001990507],"domain_scores_gemma":[0.04100872,0.7903078,0.02394458,0.1102545,0.03288364,0.001600776],"domain_codex":"methods","domain_gemma":"reproducibility","domain_candidate":"reproducibility","domain_consensus":null,"study_design_codex":"design_other","study_design_gemma":"observational","study_design_scores_codex":[0.009879072,0.001569547,0.02746289,0.03724575,0.02747286,0.001679519,0.01889499,0.04128511,0.002675622,0.3785039,0.04776548,0.4055654],"study_design_scores_gemma":[0.005865147,0.004826704,0.01172787,0.02439372,0.007809132,0.001194192,0.002528844,0.1496731,0.008738781,0.6582934,0.124074,0.0008750616],"study_design_candidate":"observational","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.01741498,0.004870389,0.9158486,0.005337106,0.004317739,0.03348407,0.002350778,0.002955659,0.01342065],"genre_scores_gemma":[0.1673325,0.001701121,0.7115328,0.002049503,0.0008276968,0.1094504,0.002199769,0.002445607,0.002460665],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.3239636,"threshold_uncertainty_score":0.3995049,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.3958113089703145,"score_gpt":0.5803904642367242,"score_spread":0.1845791552664098,"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."}}