{"id":"W3110959118","doi":"10.1093/jeea/jvaa019","title":"How Much Can We Generalize From Impact Evaluations?","year":2020,"lang":"en","type":"article","venue":"Journal of the European Economic Association","topic":"Evaluation and Performance Assessment","field":"Decision Sciences","cited_by":232,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Toronto","funders":"","keywords":"Exploit; Set (abstract data type); Sample (material); Government (linguistics); Sample size determination; Econometrics; Computer science; Treatment effect; Economics; Statistics; Mathematics; Computer security","routes":{"ca_aff":true,"ca_fund":false,"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":[],"consensus_categories":[],"category_scores_codex":[0.007320421,0.00010399,0.0002446826,0.00007574658,0.0001531224,0.0005864402,0.0007784068,0.00003212687,0.0008422736],"category_scores_gemma":[0.002802406,0.00006228925,0.0003263739,0.0001664326,0.00001355559,0.0006021761,0.00009173233,0.0002191638,0.0004135106],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0009323934,"about_ca_system_score_gemma":0.0003907502,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00001555075,"about_ca_topic_score_gemma":0.00004178996,"domain_scores_codex":[0.9962759,0.001696399,0.0007943034,0.0001561187,0.0009417768,0.0001354861],"domain_scores_gemma":[0.9961337,0.0005010912,0.002630359,0.0002393436,0.0003706891,0.000124764],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"observational","study_design_gemma":"observational","study_design_scores_codex":[0.00005804953,0.00004741075,0.4746076,0.000001661335,0.0003965255,0.000002774077,0.006470362,0.03717791,0.003145216,0.000151129,0.3823286,0.09561273],"study_design_scores_gemma":[0.001640077,0.0001571951,0.8663104,0.00001937289,0.0001161455,0.00000512399,0.001518112,0.03936847,0.0007657313,0.00882196,0.08107654,0.0002008946],"study_design_candidate":"observational","study_design_consensus":"observational","genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.8437785,0.00009874553,0.000679066,0.1496118,0.001249911,0.0001002451,0.00005649829,0.000007750177,0.004417547],"genre_scores_gemma":[0.9944611,0.00007317227,0.0002883862,0.001463131,0.00156545,4.468501e-7,0.000003498725,0.00001258333,0.002132191],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.3917027,"threshold_uncertainty_score":0.9222306,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.1830321189311144,"score_gpt":0.4270938186999413,"score_spread":0.2440616997688269,"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."}}