analysis op• 数据种类:design × signal → table
• 调用: import fullseye as fs; fs.ledger.rsm_fit(design=None, y=None, model='quadratic') -> 'dict'(要直接调用实现,import multivar; multivar.rsm_fit(design=None, y=None, model='quadratic') -> 'dict';从台账取用则 opsmultivar.get("rsm_fit"))
在 design（编码单位）上拟合响应曲面：方差分析、失拟检验、驻点。
> 以下的详细说明为原文 —— 摘要与标题已翻译。
model: "linear" (b0 + sum b_i x_i), "interaction" (+ b_ij x_i x_j) or "quadratic" (+ b_ii x_i^2) — fitted by least squares on the coded matrix of the design (so coefficients are comparable across factors).
Refuses a model the plan cannot support: if the model matrix is rank-deficient (a two-level plan cannot see x^2 — every x_i^2 equals the intercept; a resolution III fraction cannot separate a main effect from a two-factor interaction) the op raises ValueError naming the aliased terms, instead of returning one of infinitely many least-squares solutions. It also needs more runs than terms (a residual to judge with).
Returns coefficients with standard errors and t, the ANOVA (model / residual and, when the plan has replicated points, lack of fit vs pure error with its F test), R^2, adjusted R^2, predicted R^2 (from PRESS). For the quadratic model, the stationary point x_s = -B^{-1} b / 2 (B symmetric with b_ii on the diagonal and b_ij/2 off it) in coded and natural units, the predicted response there, the eigenvalues of B (canonical analysis: all negative = maximum, all positive = minimum, mixed = saddle, any near zero = ridge) and whether x_s lies inside the explored region.
• 示例数据目录(下载 URL / 许可证) —— 2-D 用 skimage.data(BSD/公有领域)加合成图,3-D 给出真实数据源(Stanford/PDS 等)的下载 URL。
• 算子来历与参考文献 —— 该算子族所依据的研究/方法出处。
• 算法的正典(作者、年份)与用途见上面的族使用指南。
• multivar_inspection_tour — py -3.11 examples/multivar_inspection_tour.py
table 作为输入)—
analysis)mv_pca · mv_pls · doe_main_effects
*Provenance: multivar.py — MULTIVAR 算子登记表。本条目由 tools/opdocs.py md 自动生成(请勿手工编辑)。*
© 2026 Kazufumi Furuse — Fullseye operator documentation. Licensed under Apache-2.0.