Struct ModelSelection 

pub struct ModelSelection;

Model selection criteria: AIC, BIC, Bayes factor, and cross-validation.

Implementations

impl ModelSelection

pub fn aic(log_likelihood: f64, n_params: usize) -> f64

Akaike Information Criterion: AIC = 2k - 2 log L.

Arguments

• log_likelihood - Maximised log-likelihood.
• n_params - Number of free parameters k.

pub fn aicc(log_likelihood: f64, n_params: usize, n_data: usize) -> f64

Corrected AIC for small samples: AICc = AIC + 2k(k+1)/(n-k-1).

pub fn bic(log_likelihood: f64, n_params: usize, n_data: usize) -> f64

Bayesian Information Criterion: BIC = k ln(n) - 2 log L.

Arguments

• log_likelihood - Maximised log-likelihood.
• n_params - Number of free parameters k.
• n_data - Number of data points n.

pub fn log_bayes_factor(log_evidence_1: f64, log_evidence_2: f64) -> f64

Bayes factor (in log scale): log BF₁₂ = log p(D|M₁) - log p(D|M₂).

pub fn jeffreys_scale(log_bf: f64) -> &'static str

Interprets the log Bayes factor according to Jeffreys’ scale.

Returns a descriptive string.

pub fn k_fold_cv_mse( x: &[Vec<f64>], y: &[f64], k: usize, alpha: f64, beta_noise: f64, ) -> f64

K-fold cross-validation mean squared error.

Arguments

• x - Feature matrix (n × d, row-major as Vec<Vecf64>).
• y - Target vector (length n).
• k - Number of folds.
• alpha - Prior precision for Bayesian linear regression.
• beta - Noise precision.

pub fn loo_cv_log_predictive( x: &[Vec<f64>], y: &[f64], alpha: f64, beta_noise: f64, ) -> f64

Pseudo Bayes factor approximation using LOO-CV log predictive density.

Trait Implementations

impl Clone for ModelSelection

fn clone(&self) -> ModelSelection

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for ModelSelection

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

Formats the value using the given formatter.