mt opmatrix → tableimport fullseye as fs; fs.ledger.spc_mt_distance(data, mean=None, std=None, inv_corr=None, threshold=3.0) (実装を直接呼ぶなら import spc; spc.spc_mt_distance(data, mean=None, std=None, inv_corr=None, threshold=3.0)、台帳から引くなら opsspc.get("spc_mt_distance"))Score observations against an MT-method unit space.
data is a 2-D array of shape (m, p). Pass the mean, std and
inv_corr from :func:spc_mt_unit_space to score new product against a
stored unit space; omit all three and the unit space is built from data
itself, which is the training case and makes md_sq_mean land on
(m-1)/m.
Each row’s distance is MD_i = sqrt(z_i' R^-1 z_i / p) with
z_i = (x_i - mean) / std. Rows with MD > threshold are reported in
flagged; the rest in inside. Dividing the quadratic form by p is
what puts the unit space at 1 regardless of how many features are used, so a
threshold carries over when features are added or removed.
Returns a dict with the per-row md and md_sq arrays, flagged,
inside, the threshold, md_sq_mean, and n/p.
Relationship to :func:spc_hotelling_t2 (used as the oracle in the tests):
on the standardised matrix the covariance is the correlation, so
md**2 == t2 / p exactly for the same data.
Raises ValueError: a non-2-D / empty data, a non-positive
threshold, a unit space given only in part, a mean / std /
inv_corr whose shape disagrees with p, a non-positive entry in
std, or non-finite input.
py -3.11 examples/poc_mt_hidden_fault.pypy -3.11 examples/poc_spc.pytable を入力に取れる)msa_anova_table · msa_gauge_rr · msa_bias_linearity · msa_attribute_agreement · gum_standard_uncertainty · gum_propagate · gum_expanded · gum_monte_carlo
mt)spc_mt_unit_space · spc_mt_sn_ratio
Provenance: spc.py — SPC 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.