record_id,quickpls_version,fixture,cell,native,reference,abs_error,relative_error,passed,evidence_kind,convention_alignment
prediction-mean-phase2,Retained training-mean v4 prediction witness,prediction-mean/holdout,holdout/aggregate_cases/0/mean_ia_squared_loss,0.1598254761528995,0.1598254761528995,0.0,0.0,True,external-library-match + independently-implemented-math,Fresh cSEM normalized training weights; statsmodels OLS; independent training-only preparation/held-out masks/case aggregation and SciPy Student-t
prediction-mean-phase2,Retained training-mean v4 prediction witness,prediction-mean/holdout,holdout/aggregate_cases/0/mean_lm_squared_loss,0.3038682941098941,0.303868294109894,5.551115123125783e-17,1.8268161669799682e-16,True,external-library-match + independently-implemented-math,Fresh cSEM normalized training weights; statsmodels OLS; independent training-only preparation/held-out masks/case aggregation and SciPy Student-t
prediction-mean-phase2,Retained training-mean v4 prediction witness,prediction-mean/holdout,holdout/aggregate_cases/0/mean_squared_loss,0.31479713250214364,0.3147971325021685,2.4868995751603507e-14,7.90000707882312e-14,True,external-library-match + independently-implemented-math,Fresh cSEM normalized training weights; statsmodels OLS; independent training-only preparation/held-out masks/case aggregation and SciPy Student-t
prediction-mean-phase2,Retained training-mean v4 prediction witness,prediction-mean/holdout,holdout/aggregate_cases/1/mean_ia_squared_loss,0.6125946301910633,0.6125946301910633,0.0,0.0,True,external-library-match + independently-implemented-math,Fresh cSEM normalized training weights; statsmodels OLS; independent training-only preparation/held-out masks/case aggregation and SciPy Student-t
prediction-mean-phase2,Retained training-mean v4 prediction witness,prediction-mean/holdout,holdout/aggregate_cases/1/mean_lm_squared_loss,0.13799350615661515,0.13799350615661557,4.163336342344337e-16,3.0170523659418896e-15,True,external-library-match + independently-implemented-math,Fresh cSEM normalized training weights; statsmodels OLS; independent training-only preparation/held-out masks/case aggregation and SciPy Student-t
prediction-mean-phase2,Retained training-mean v4 prediction witness,prediction-mean/holdout,holdout/aggregate_cases/1/mean_squared_loss,0.13013288316017724,0.1301328831601544,2.2842838731662596e-14,1.755347163372223e-13,True,external-library-match + independently-implemented-math,Fresh cSEM normalized training weights; statsmodels OLS; independent training-only preparation/held-out masks/case aggregation and SciPy Student-t