{"id":"W3208880666","doi":"10.5539/jel.v10n6p103","title":"Designing Standards-Setting for Levels of Mathematical Proficiency in Measurement and Geometry: Multidimensional Item Response Model","year":2021,"lang":"en","type":"article","venue":"Journal of Education and Learning","topic":"Mathematics Education and Teaching Techniques","field":"Social Sciences","cited_by":5,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"National Research Council of Thailand; Khon Kaen University","keywords":"Mathematical model; Set (abstract data type); Reliability (semiconductor); Quality (philosophy); Computer science; Process (computing); Construct (python library); Mathematics; Statistics","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":"codex-gemma-dda1882f352a","candidate_categories":["metaresearch"],"consensus_categories":[],"category_scores_codex":[0.01191016,0.00005123621,0.0001516744,0.000195987,0.0002104868,0.00003822934,0.00004256454,0.00003764621,0.00002556008],"category_scores_gemma":[0.01777921,0.00004719893,0.00003079006,0.0001374548,0.00004856266,0.0001082173,0.00001319614,0.0001916461,5.919233e-8],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0001246121,"about_ca_system_score_gemma":0.001840547,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.000005318275,"about_ca_topic_score_gemma":0.00000438573,"domain_scores_codex":[0.998584,0.0004491054,0.0003664357,0.00007869336,0.000422889,0.00009882826],"domain_scores_gemma":[0.9979352,0.0008164236,0.0003214887,0.00003941462,0.0008184321,0.00006899601],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"qualitative","study_design_gemma":"qualitative","study_design_scores_codex":[0.0004115334,0.002075213,0.02142757,0.00105598,0.00007600024,0.000003154228,0.5263968,0.0008596728,0.1228458,0.1992953,0.00118985,0.1243632],"study_design_scores_gemma":[0.003903393,0.001054642,0.05997531,0.00933861,0.0002452593,0.0002165814,0.5886691,0.06607271,0.03789506,0.2152431,0.01623218,0.001154082],"study_design_candidate":"qualitative","study_design_consensus":"qualitative","genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.7979146,0.0003346605,0.1997199,0.00142426,0.00006354372,0.0001300983,0.000001041148,0.000006956636,0.0004050022],"genre_scores_gemma":[0.8051722,0.00002060388,0.1945602,0.00003217999,0.00002685876,0.00000415421,1.88686e-7,0.000004404697,0.0001791292],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.1232091,"threshold_uncertainty_score":0.9904944,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.08451347358686771,"score_gpt":0.4060367442683351,"score_spread":0.3215232706814674,"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."}}