Struct StandardizationInfo 

pub struct StandardizationInfo {
    pub x_mean: Vec<f64>,
    pub x_scale: Vec<f64>,
    pub column_squared_norms: Vec<f64>,
    pub y_mean: f64,
    pub y_scale: Option<f64>,
    pub y_scale_before_sqrt_weights_normalized: Option<f64>,
    pub intercept: bool,
    pub standardized_x: bool,
    pub standardized_y: bool,
}

Information stored during standardization, used to unstandardize coefficients.

This struct captures all the information needed to transform coefficients from the standardized space back to the original data scale.

Fields

• x_mean - Mean of each predictor column (length p)
• x_scale - Scale factor for each predictor column (length p)
• y_mean - Mean of response variable
• y_scale - Scale factor for response (optional, used for lambda path)
• intercept - Whether an intercept term was included
• standardized_x - Whether X was standardized
• standardized_y - Whether y was standardized

Example

let info = StandardizationInfo {
    x_mean: vec![0.0, 5.0],
    x_scale: vec![1.0, 2.0],
    column_squared_norms: vec![1.0, 1.0],
    y_mean: 10.0,
    y_scale: Some(3.0),
    y_scale_before_sqrt_weights_normalized: Some(3.0),
    intercept: true,
    standardized_x: true,
    standardized_y: false,
};

assert_eq!(info.x_mean.len(), 2);
assert!(info.intercept);

Fields

x_mean: Vec<f64>

Mean of each predictor column

x_scale: Vec<f64>

Scale factor for each predictor column

column_squared_norms: Vec<f64>

Squared norm of each predictor column after standardization. This is used in the coordinate descent update denominator.

• With intercept and standardize: column_squared_norms[j] = 1.0 (unit norm after centering)
• Without intercept and standardize: column_squared_norms[j] = 1.0 + x_squared_mean/x_centered_variance (glmnet formula)
• Without standardize: column_squared_norms[j] = ||x_j||^2 (actual squared norm)

y_mean: f64

Mean of response variable

y_scale: Option<f64>

Scale factor for response (for lambda path construction) This is the norm AFTER sqrt_weights_normalized transformation and centering: sqrt(sum((sqrt_weights_normalized*(y-ym))^2))

y_scale_before_sqrt_weights_normalized: Option<f64>

Scale factor for response BEFORE sqrt_weights_normalized transformation: sqrt(sum((y-ym)^2)) This is used for lambda scaling between original and standardized data

intercept: bool

Whether an intercept was included

standardized_x: bool

Whether X was standardized

standardized_y: bool

Whether y was standardized

Trait Implementations

impl Clone for StandardizationInfo

fn clone(&self) -> StandardizationInfo

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for StandardizationInfo

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

Formats the value using the given formatter.