{"id":"W4411245502","doi":"10.1111/test.12410","title":"Uncovering key statistical concepts from an investigation of four standardized tests","year":2025,"lang":"en","type":"article","venue":"Teaching Statistics","topic":"Statistics Education and Methodologies","field":"Mathematics","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Ottawa","funders":"Social Sciences and Humanities Research Council of Canada; University of Ottawa","keywords":"Key (lock); Statistical hypothesis testing; Computer science; Standardized test; Statistical analysis; Statistics; Econometrics; Mathematics education; Psychology; Mathematics","routes":{"ca_aff":true,"ca_fund":true,"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":[0.02927185,0.0009971219,0.0006499558,0.006990712,0.0009798066,0.003327148,0.001471326,0.0008074893,0.00148636],"category_scores_gemma":[0.1717215,0.0004012507,0.0009204903,0.005101004,0.003987916,0.002990845,0.003242867,0.001829586,0.0002884491],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.002478365,"about_ca_system_score_gemma":0.004934256,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.001192631,"about_ca_topic_score_gemma":0.001518382,"domain_scores_codex":[0.9642224,0.0167954,0.00527255,0.001506026,0.01108738,0.001116256],"domain_scores_gemma":[0.7428171,0.1912771,0.02440938,0.008347129,0.03129354,0.001855794],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"observational","study_design_gemma":"observational","study_design_scores_codex":[0.0006690449,0.00148487,0.4461691,0.001946935,0.0002953116,0.001294909,0.07774322,0.003612731,0.01902606,0.0284391,0.002271618,0.417047],"study_design_scores_gemma":[0.000170605,0.00676676,0.701795,0.002121335,0.0004568968,0.002007289,0.1120279,0.03061618,0.06174142,0.04490677,0.03700702,0.0003828483],"study_design_candidate":"observational","study_design_consensus":"observational","genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.9622501,0.0002551322,0.03063329,0.0003402971,0.00006472031,0.0007007825,0.0001879529,0.0001121964,0.005455545],"genre_scores_gemma":[0.9715134,0.0001702438,0.0264596,0.00006911802,0.00002192905,0.0006805279,0.0002814181,0.00003213429,0.0007717446],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.02927185,"threshold_uncertainty_score":0.1548062,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.1264707516143293,"score_gpt":0.465697483722037,"score_spread":0.3392267321077077,"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."}}