Struct QuantileConfig 

pub struct QuantileConfig {
    pub levels: Vec<f64>,
    pub horizon: usize,
    pub initial_window: usize,
    pub step: usize,
    pub strategy: CalibrationStrategy,
    pub conformal: bool,
    pub lower_bound: Option<f64>,
    pub upper_bound: Option<f64>,
}

Configuration for QuantileForecaster.

Fields

levels: Vec<f64>

Quantile levels to produce, each in (0, 1).

horizon: usize

Maximum horizon to calibrate. Residuals are bucketed per step 1..=horizon.

initial_window: usize

Number of observations in the initial training window.

step: usize

Observations to advance between rolling origins.

strategy: CalibrationStrategy

How to advance the model between origins.

conformal: bool

Apply the finite-sample conformal rank correction.

lower_bound: Option<f64>

Optional hard lower bound on predicted quantiles.

Residual quantiles are additive offsets on the point forecast, so they can cross a physical limit — a PV series bounded at zero can still get a negative lower bound when the point forecast is noisy. Learning the residual shape from data and stating the domain explicitly are different jobs; this is the second one.

upper_bound: Option<f64>

Optional hard upper bound on predicted quantiles (e.g. inverter nameplate capacity, 100% state of charge).

Implementations

impl QuantileConfig

pub fn with_bounds(self, lo: Option<f64>, hi: Option<f64>) -> Self

Constrain predicted quantiles to [lo, hi].

pub fn central(level: f64, horizon: usize) -> Self

Config producing a symmetric central interval at level coverage (e.g. 0.9 → the 5% and 95% quantiles plus the median).

Trait Implementations

impl Clone for QuantileConfig

fn clone(&self) -> QuantileConfig

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for QuantileConfig

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

Formats the value using the given formatter.

impl Default for QuantileConfig

fn default() -> Self

Returns the “default value” for a type.