Struct ScatterFit 

pub struct ScatterFit {
    pub location: Array1<f64>,
    pub scatter: Array2<f64>,
    pub distances: Array1<f64>,
    pub weights: Array1<f64>,
}

A fitted robust location–scatter estimate.

The shared result type of MScatter and Ogk. FAST-MCD carries extra structure (a raw estimate and the retained subset), so it returns its own McdFit, but that type implements RobustScatter too, so all three are interchangeable behind the trait and the Mahalanobis/outlier map is written once against (μ̂, Σ̂). Tyler does not use this type: it returns a bespoke TylerFit implementing no shared trait, because its shape-only distances are not χ²-calibrated, so this type’s Gaussian outlier map would misapply to them.

Fields

location: Array1<f64>

Robust location estimate μ̂ (a p-vector).

scatter: Array2<f64>

Robust scatter (covariance) estimate Σ̂ (p × p, symmetric positive definite).

distances: Array1<f64>

Robust Mahalanobis distances dᵢ of the rows the fit was computed on.

weights: Array1<f64>

Final per-observation weights (their meaning is estimator-specific: a hard 0/1 acceptance for a reweighting estimator, a smooth w(dᵢ) for an M-estimator).

Trait Implementations

impl Clone for ScatterFit

fn clone(&self) -> ScatterFit

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for ScatterFit

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl RobustScatter for ScatterFit

fn location(&self) -> &Array1<f64>

Robust location estimate μ̂.

fn scatter(&self) -> &Array2<f64>

Robust scatter estimate Σ̂.

fn distances(&self) -> &Array1<f64>

Robust Mahalanobis distances of the fitted rows.

fn distance_cutoff(&self, quantile: f64) -> f64

The distance cutoff √χ²_{p, quantile} (e.g. quantile = 0.975) against which robust Mahalanobis distances are compared to flag outliers.

fn outliers(&self, quantile: f64) -> Vec<bool>

Flag each fitted observation as an outlier when its robust distance exceeds distance_cutoff(quantile). Under a p-variate Gaussian the squared distances are χ²_p, so quantile = 0.975 flags ≈ 2.5% of clean data.