{"id":"W4407927480","doi":"10.3102/10769986251314833","title":"Using the Information Metric to Analyze Clinical Rating Scales","year":2025,"lang":"en","type":"article","venue":"Journal of Educational and Behavioral Statistics","topic":"Health Systems, Economic Evaluations, Quality of Life","field":"Economics, Econometrics and Finance","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"Dalhousie University; University of Manitoba; Ottawa Hospital; McGill University","funders":"Vetenskapsrådet","keywords":"Rating scale; Metric (unit); Item response theory; Statistics; Computer science; Econometrics; Mathematics; Psychology; Psychometrics","routes":{"ca_aff":true,"ca_fund":false,"ca_venue":false,"about_ca":false,"invisible_to_affiliation_only":false},"retraction":null,"screen":null,"direct_labels":[{"model":"gpt","categories":[],"domain":null,"study_design":"design_other","genre":"methods","about_ca_system":false,"about_ca_topic":false,"confidence":"high","status":"direct model label, unvalidated"},{"model":"grok","categories":[],"domain":null,"study_design":"design_other","genre":"methods","about_ca_system":false,"about_ca_topic":false,"confidence":"high","status":"direct model label, unvalidated"},{"model":"opus","categories":[],"domain":null,"study_design":"simulation_or_modeling","genre":"methods","about_ca_system":false,"about_ca_topic":false,"confidence":"high","status":"direct model label, unvalidated"}],"prediction":{"model_version":"codex-gemma-dda1882f352a","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.006545846,0.0000808586,0.0003917511,0.0004038088,0.0002145503,0.0001341236,0.0001491837,0.00005301057,0.0001172389],"category_scores_gemma":[0.004419304,0.00007183726,0.00006172382,0.0003323689,0.00006012637,0.0004193679,0.00003241048,0.0001714825,0.00004145177],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0002035183,"about_ca_system_score_gemma":0.000356725,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0002103465,"about_ca_topic_score_gemma":0.00004132719,"domain_scores_codex":[0.9963011,0.0001457809,0.003234408,0.00009722773,0.00009543337,0.0001260399],"domain_scores_gemma":[0.9960326,0.001593832,0.001842987,0.0001080844,0.0003062389,0.0001162685],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"observational","study_design_gemma":"observational","study_design_scores_codex":[0.00002148489,0.000207004,0.6661284,0.0000810518,0.00006178994,3.120941e-7,0.000768404,0.0004539726,0.00000215293,0.2540155,0.07274819,0.005511687],"study_design_scores_gemma":[0.0005732139,0.000176201,0.9102278,0.0001096783,0.00007547694,0.00002309127,0.002167178,0.003910767,0.000002062404,0.04766378,0.03486928,0.0002014538],"study_design_candidate":"observational","study_design_consensus":"observational","genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.7871441,0.001201507,0.1640036,0.04461309,0.0019636,0.0003119374,0.0004180362,0.000003328634,0.0003408355],"genre_scores_gemma":[0.8548202,0.00009516462,0.1366356,0.007537296,0.0005678709,0.000008666315,0.00002594625,0.000007193496,0.0003020895],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.2440994,"threshold_uncertainty_score":0.5290639,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.4986897028064809,"score_gpt":0.5681965655826773,"score_spread":0.06950686277619639,"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."}}