Struct Ordination 

pub struct Ordination {
    pub sample_ids: Vec<String>,
    pub species_ids: Vec<String>,
    pub constraint_ids: Vec<String>,
    pub eigvals: Vec<f64>,
    pub proportion_explained: Vec<f64>,
    pub sample_scores: Vec<f64>,
    pub species_scores: Vec<f64>,
    pub biplot_scores: Vec<f64>,
    pub biplot_axes: usize,
    pub sample_constraints: Vec<f64>,
}

Result of a Redundancy Analysis. Eigenvalues, proportion explained, and the site/species scores plus biplot and site-constraint scores follow skbio.stats.ordination.rda. The first n_constrained axes are canonical (constrained by the explanatory variables); the rest are the unconstrained PCA of the residuals.

Fields

sample_ids: Vec<String>

species_ids: Vec<String>

constraint_ids: Vec<String>

eigvals: Vec<f64>

proportion_explained: Vec<f64>

sample_scores: Vec<f64>

Row-major n_samples × n_axes.

species_scores: Vec<f64>

Row-major n_species × n_axes.

biplot_scores: Vec<f64>

Biplot scores follow the left singular vectors of the fitted values, so they span only the constrained axes: row-major n_constraints × biplot_axes.

biplot_axes: usize

sample_constraints: Vec<f64>

Row-major n_samples × n_axes.

Implementations

impl Ordination

pub fn compute( response: &Matrix, constraints: &Matrix, scaling: u8, scale_y: bool, ) -> Result<Ordination>

RDA per Legendre & Legendre 1998 §11.1: regress centred y on centred x (SVD least squares), SVD the fitted values for the canonical axes, SVD the residuals for the unconstrained axes, then apply scaling 1 or 2.

Errors

Errors when the two tables disagree on sample count or when x has more columns than rows (an under-determined regression), matching skbio.

pub fn write_tsv<W: Write>(&self, out: W) -> Result<()>

Write the ordination as a flat TSV with # eigenvalues, # samples, # species, # biplot, and # site_constraints blocks, axes RDA1...

Errors

Propagates write errors.