analysis opdesign × signal → tableimport fullseye as fs; fs.ledger.doe_main_effects(design=None, y=None, goal='larger') -> 'dict' (実装を直接呼ぶなら import multivar; multivar.doe_main_effects(design=None, y=None, goal='larger') -> 'dict'、台帳から引くなら opsmultivar.get("doe_main_effects"))Level means per factor — the response table of a designed experiment — with dispersion and SN.
y is (n,) or (n, r) with r repeated measurements (or noise-array conditions) per
run. For each factor and level: the mean of y (location), the mean of log s over the
runs at that level (dispersion, needs r >= 2), the main-effect range (max - min of the
level means) and its share of the total sum of squares.
SN ratios are computed per run (Taguchi): larger-the-better
-10 log10(mean(1/y^2)) (the same formula as spc.spc_mt_sn_ratio, reused),
smaller-the-better -10 log10(mean(y^2)), nominal-the-best 10 log10(ybar^2/s^2),
and their level means are reported in sn. They are not the basis of the
recommended levels: an SN ratio mixes location and dispersion in one number, and a
level can raise it by lowering the mean (Box 1988; Nair 1992). best_levels chooses
from the location table by goal ("larger", "smaller", or "nominal" —
which picks the levels of smallest dispersion and leaves the mean to an adjustment
factor), and sn_agrees says where the SN table would have chosen differently.
py -3.11 examples/multivar_inspection_tour.pytable を入力に取れる)—
analysis)Provenance: multivar.py — MULTIVAR 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.