First release.
nutrigon_fit() fits a quadratic response surface and
locates the nutritional optimum with a hybrid estimator: the closed-form
stationary point when it is admissible (inside the design hull, correct
curvature sign), otherwise a constrained Monte Carlo search summarised
by the point with the highest (or lowest) fitted value.
model_type selects ordinary least squares
("lm"), a generalized linear model with any family
including quasi-families ("glm"; optimum located on the
link scale), or a linear mixed-effects model ("lmer";
population-level optimum; requires lme4 and lmerTest).nutrigon_coefs(), nutrigon_elongation(),
nutrigon_snr() and nutrigon_curvature() return
the fitted quadratic coefficients, the curvature elongation ratio, the
signal-to-noise ratio, and local Gaussian/mean curvature with a
surface-area ratio.nutrigon_compare() estimates the angular separation
between two optima with a percentile bootstrap interval that refits both
surfaces in every replicate. Resampling is by case (default) or
stratified within design points, paired or unpaired. Returns the
locations of the optima, each surface’s angular support, the bootstrap
distribution and the Monte Carlo fallback rate.nutrigon_matrix() builds the N x N angular trade-off
matrix with joint bootstrap uncertainty (the same replicate of each
surface is used in every pair) and the bootstrap stability of a
hierarchical clustering.nutrigon_equivalence() tests whether two optima are
aligned within a pre-specified angle.nutrigon_build_grid() simulates a validation grid,
either on the user’s own design and at the signal-to-noise ratio of the
fitted surfaces (recommended) or on a generic rail design.nutrigon_check() estimates the coverage expected for
one result from the nearest grid configurations, with a Wilson interval,
excluding near-zero angles and warning when the grid comes from a
different design.nutrigon_grid(), nutrigon_hull() and
nutrigon_region_polygon() prepare fitted surfaces for
ggplot2.lee_drosophila, zanco_survival,
zanco_wingsize, zanco_tradeoff_matrix and
nutrigon_validation_grid.