{"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":"codex-gemma-dda1882f352a","candidate_categories":["metaresearch"],"consensus_categories":[],"category_scores_codex":[0.001418818,0.0002259426,0.0004969591,0.0001460735,0.0001807781,0.00007079393,0.0002441628,0.0001069608,0.0001583641],"category_scores_gemma":[0.01809757,0.0002254825,0.00003342056,0.0001328285,0.0002442015,0.0001024026,0.00007527111,0.0003746859,0.000006473908],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0001322417,"about_ca_system_score_gemma":0.0004718862,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0004265375,"about_ca_topic_score_gemma":0.00009673025,"domain_scores_codex":[0.9974295,0.0009084835,0.0006994226,0.0003387009,0.0003617754,0.0002620738],"domain_scores_gemma":[0.9869681,0.01189543,0.0003205368,0.0004538051,0.0002409251,0.0001211864],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"theoretical_or_conceptual","study_design_gemma":"theoretical_or_conceptual","study_design_scores_codex":[0.00007298918,0.00008254484,0.008893605,0.0002012976,0.00009614635,0.000008320088,0.004142117,0.00004521419,0.001698368,0.925939,0.04161719,0.01720325],"study_design_scores_gemma":[0.0007625777,0.00009559357,0.02857159,0.0001518124,0.0001695393,0.000001550915,0.001059257,0.001399334,0.00126371,0.9650363,0.00126228,0.0002264175],"study_design_candidate":"theoretical_or_conceptual","study_design_consensus":"theoretical_or_conceptual","genre_codex":"methods","genre_gemma":"methods","genre_scores_codex":[0.1523105,0.00002593102,0.8410498,0.000098207,0.0007795753,0.000214777,0.004353998,0.0001406241,0.001026607],"genre_scores_gemma":[0.2445582,0.000007674013,0.7544042,0.0001604783,0.00005612599,0.0000132384,0.000315566,0.00002472383,0.0004597983],"genre_candidate":"methods","genre_consensus":"methods","teacher_disagreement_score":0.09224768,"threshold_uncertainty_score":0.9901734,"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."}}