{"id":"W2153720865","doi":"10.55016/ojs/ajer.v49i3.54984","title":"Which Model is Best? Robustness Properties to Justify Model Choice Among Unidimensional IRT Models under Item Parameter Drift","year":2003,"lang":"en","type":"article","venue":"Alberta Journal of Educational Research","topic":"Advanced Causal Inference Techniques","field":"Mathematics","cited_by":21,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"","keywords":"Robustness (evolution); Econometrics; Item response theory; Statistical physics; Computer science; Mathematics; Statistics; Psychometrics; Physics","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":"metacan-v3-hybrid-931329e0061c","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.09368901,0.002083157,0.003568962,0.003560818,0.001312823,0.004802101,0.003274217,0.004834429,0.0047225],"category_scores_gemma":[0.3423345,0.001163575,0.004077361,0.00293077,0.003860837,0.009890272,0.004481031,0.006152079,0.000735282],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001969751,"about_ca_system_score_gemma":0.00194994,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.002531211,"about_ca_topic_score_gemma":0.001558293,"domain_scores_codex":[0.9506567,0.03823514,0.002170407,0.005629913,0.002319317,0.0009884901],"domain_scores_gemma":[0.5935737,0.3584717,0.01664263,0.02674463,0.003165794,0.001401476],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"theoretical_or_conceptual","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.002633548,0.0003932352,0.03200592,0.0009609514,0.005220399,0.000810069,0.00186773,0.3298897,0.002357385,0.5197732,0.003082019,0.1010058],"study_design_scores_gemma":[0.0002064124,0.0003058951,0.003027221,0.000152213,0.0004076876,0.0002521146,0.0002157764,0.542392,0.0009712331,0.4508078,0.001164333,0.00009734921],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.1002552,0.001111621,0.8868089,0.006093733,0.0001606744,0.0003467185,0.0005738188,0.0005105585,0.004138859],"genre_scores_gemma":[0.8385682,0.0006220308,0.1576775,0.0009956186,0.0003307009,0.0004119621,0.000587895,0.0002101519,0.0005959765],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.09368901,"threshold_uncertainty_score":0.4954808,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.4434345342535627,"score_gpt":0.4856145623288259,"score_spread":0.04218002807526317,"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."}}