Struct RegressionFit 

pub struct RegressionFit {
    pub coefficients: Array1<f64>,
    pub scale: Scale,
    pub residuals: Array1<f64>,
    pub weights: Array1<f64>,
    pub rho: Box<dyn RhoFunction>,
    pub breakdown_point: f64,
}

A fitted robust regression, carrying enough state to report the theory.

Fields

coefficients: Array1<f64>

Estimated coefficients.

scale: Scale

Estimated residual scale.

residuals: Array1<f64>

Residuals y − Xβ̂.

weights: Array1<f64>

Final IRLS weights.

rho: Box<dyn RhoFunction>

The loss used, retained so influence/efficiency can be reported. This is the estimator’s final-stage loss (e.g. the M-step biweight for MM), so the reported efficiency and covariance describe the delivered fit.

breakdown_point: f64

Asymptotic breakdown point ε*, carried as data set by the producing estimator, not derived from rho, since ρ alone does not determine an estimator’s breakdown. Plain M-regression sets 0.0 (leverage); S/MM set min(δ/ρ_sup, 1 − δ/ρ_sup).

Trait Implementations

impl RobustEstimator for RegressionFit

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

Estimated coefficients.

fn scale(&self) -> Scale

Estimated residual scale.

fn influence_function(&self) -> Box<dyn Fn(f64) -> f64 + '_>

The influence function x ↦ ψ(x)/E[ψ'].

fn asymptotic_variance(&self) -> f64

Asymptotic variance E[ψ²]/(E[ψ'])².

fn gaussian_efficiency(&self) -> f64

Efficiency relative to the Gaussian MLE.

fn coef_covariance(&self, x: &Array2<f64>) -> Array2<f64>

Approximate coefficient covariance ŝ²·V·(XᵀX)⁻¹.

fn breakdown_point(&self) -> f64

Asymptotic breakdown point.