Struct PeerResult 

#[non_exhaustive]
pub struct PeerResult {
    pub beta: Vec<f64>,
    pub intercept: f64,
    pub w_bar: Vec<f64>,
    pub fitted_values: Vec<f64>,
    pub effective_df: f64,
    pub lambda: f64,
    pub penalty_type: PeerPenalty,
    pub gcv: Option<f64>,
    pub lambda_method: LambdaMethod,
}

Result of the peer estimator.

Carries the estimated coefficient function β(t), model diagnostics, and the penalty configuration used — all on the argvals grid.

Fields (Non-exhaustive)

Non-exhaustive structs could have additional fields added in future. Therefore, non-exhaustive structs cannot be constructed in external crates using the traditional Struct { .. } syntax; cannot be matched against without a wildcard ..; and struct update syntax will not work.

beta: Vec<f64>

Estimated coefficient function β(t), length m (on the argvals grid).

intercept: f64

Intercept (= ȳ, the mean of the response vector).

This is the centered-response mean, not the out-of-sample prediction intercept. Prediction on a new curve x* uses ȳ + Σ_j (x*[j]·w[j] − w_bar[j])·β[j] = (ȳ − w_bar·β) + Σ_j x*[j]·w[j]·β[j], so a predictor must combine intercept with w_bar and beta. Storing w_bar here keeps that reconstruction exact (consumed by out-of-sample prediction in a later phase).

w_bar: Vec<f64>

Column means of the Simpson-weighted design W[i,j] = data[(i,j)]·w[j], length m. Retained so out-of-sample prediction can reproduce the same centering the fit used (see intercept).

fitted_values: Vec<f64>

Fitted values ŷ_i, length n.

effective_df: f64

Effective degrees of freedom tr(H) = tr((W_c’W_c + λQ)^{-1} W_c’W_c).

lambda: f64

Smoothing parameter λ that was used.

penalty_type: PeerPenalty

Penalty family that was used.

gcv: Option<f64>

GCV score at the selected λ. Some(score) when LambdaMethod::Gcv ran; None when LambdaMethod::Fixed or LambdaMethod::Reml.

lambda_method: LambdaMethod

Which λ-selection path ran.

Implementations

impl PeerResult

pub fn predict( &self, new_data: &FdMatrix, argvals: &[f64], ) -> Result<Vec<f64>, FdarError>

Predict scalar responses for new functional observations.

Applies the fitted PEER coefficient function β(t) to new_data via ŷ*[i] = (intercept − w_bar·β) + Σ_j x*[i,j] · w[j] · β[j] where w = simpsons_weights(argvals).

Self-consistency: re-passing the training data and training argvals reproduces the training fitted_values to within floating-point rounding (≤ 1e-9 in absolute error per observation).

Errors

Returns FdarError::InvalidDimension when new_data.ncols() or argvals.len() does not equal the number of training grid points m.

Trait Implementations

impl Clone for PeerResult

fn clone(&self) -> PeerResult

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for PeerResult

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

Formats the value using the given formatter.

impl PartialEq for PeerResult

fn eq(&self, other: &PeerResult) -> bool

Equality operator ==.

fn ne(&self, other: &Rhs) -> bool

Inequality operator !=.

impl StructuralPartialEq for PeerResult