Struct HuberScale 

pub struct HuberScale {
    pub d: f64,
    pub tol: f64,
    pub maxiter: usize,
}

Huber’s “proposal 2” scaling for the IRLS weights (HuberScale).

This is the scale_est=HuberScale() option in the reference. It solves, by fixed-point iteration,

scale_{i+1}² = (1 / (n·h)) · Σ χ(r / scale_i) · scale_i²

with χ(x) = x²/2 for |x| < d and d²/2 otherwise, and the consistency constant h = (df_resid / n) · (d² + (1 − d²)·Φ(d) − 1/2 − d·φ(d)).

Fields

d: f64

Tuning constant d (default 2.5).

tol: f64

Convergence tolerance on successive scale estimates.

maxiter: usize

Maximum number of fixed-point iterations.

Implementations

impl HuberScale

pub fn scale(&self, df_resid: f64, nobs: f64, resid: &[f64]) -> f64

Evaluate Huber’s proposal-2 scale for the given residuals.

df_resid and nobs are the model degrees of freedom and the number of observations; the iteration is seeded with the MAD of resid.

Trait Implementations

impl Clone for HuberScale

fn clone(&self) -> HuberScale

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Copy for HuberScale

impl Debug for HuberScale

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

Formats the value using the given formatter.

impl Default for HuberScale

fn default() -> Self

Returns the “default value” for a type.