check opcalresult → calresultimport fullseye as fs; fs.ledger.simcal_identifiability(result=None, at=None, profile=True, level=0.95, threshold=None) -> 'dict' (実装を直接呼ぶなら import simcalcheck; simcalcheck.simcal_identifiability(result=None, at=None, profile=True, level=0.95, threshold=None) -> 'dict'、台帳から引くなら opssimcal.get("simcal_identifiability"))Fisher information, Cramér-Rao bound, correlations, collinearity and profile likelihood of every fitted knob.
Computed at the fit’s end point (or at at = {knob: value}, e.g. the true values of
a synthetic study, without touching the estimates). Returned in identifiability:
fisher (natural units), crlb (sqrt of the diagonal of its inverse — the smallest
standard deviation any unbiased estimator can have under the noise model), condition
(of the Fisher matrix on the unit-scaled knobs; > 1e6 adds a warning), correlation,
gamma_full and confounded_groups (sensitivity.mv_confounding), and with
profile=True a profile per knob: the likelihood-ratio interval where chi^2 stays
within chi^2_1(level) of its minimum (3.84 at 95 %), each other knob re-fitted at every
step (Raue et al. 2009). Knobs whose profile does not rise above the threshold on both sides
inside the range lose their point estimate: upper_only / lower_only (only one limit
exists) or undetermined. Each profile keeps its trace (knob value, chi^2 rise) of the
evaluated points. No noise model: refused.
py -3.11 examples/poc_simcal_pixel_shift.pypy -3.11 examples/poc_simcal_telecentric.pypy -3.11 examples/poc_simcal_thin_film.pypy -3.11 examples/simcal_checks_tour.pycalresult を入力に取れる)simcal_uncertainty · simcal_validate · simcal_discrepancy · simcal_report
check)simcal_uncertainty · simcal_validate · simcal_discrepancy
Provenance: simcalcheck.py — SIMCAL operator registry. この per-op ノートは tools/opdocs.py md が自動生成(手編集しない)。
© 2026 Kazufumi Furuse — Fullseye operator documentation. Licensed under Apache-2.0.