Struct InformationCriterionCalculator 

pub struct InformationCriterionCalculator {
    pub use_bias_correction: bool,
    pub calculate_weights: bool,
}

Information criterion calculator

Fields

use_bias_correction: bool

Whether to use bias correction for finite samples

calculate_weights: bool

Whether to calculate model weights

Implementations

impl InformationCriterionCalculator

pub fn new() -> Self

Create new calculator

pub fn aic( &self, log_likelihood: f64, n_parameters: usize, n_data_points: usize, ) -> InformationCriterionResult

Calculate AIC (Akaike Information Criterion) AIC = 2k - 2ln(L)

pub fn aicc( &self, log_likelihood: f64, n_parameters: usize, n_data_points: usize, ) -> Result<InformationCriterionResult>

Calculate AICc (Corrected AIC for finite samples) AICc = AIC + 2k(k+1)/(n-k-1)

pub fn bic( &self, log_likelihood: f64, n_parameters: usize, n_data_points: usize, ) -> InformationCriterionResult

Calculate BIC (Bayesian Information Criterion) BIC = k*ln(n) - 2ln(L)

pub fn dic( &self, log_likelihood_mean: f64, log_likelihood_samples: &[f64], n_data_points: usize, ) -> Result<InformationCriterionResult>

Calculate DIC (Deviance Information Criterion) DIC = D(θ̄) + 2p_D, where p_D is effective number of parameters

pub fn waic( &self, pointwise_log_likelihoods: &Array2<f64>, ) -> Result<InformationCriterionResult>

Calculate WAIC (Watanabe-Akaike Information Criterion) WAIC = -2 * (lppd - p_WAIC)

pub fn looic( &self, pointwise_log_likelihoods: &Array2<f64>, pareto_k_diagnostics: Option<&Array1<f64>>, ) -> Result<InformationCriterionResult>

Calculate LOOIC (Leave-One-Out Information Criterion) using Pareto smoothed importance sampling

pub fn tic( &self, log_likelihood: f64, fisher_information_trace: f64, n_data_points: usize, ) -> Result<InformationCriterionResult>

Calculate TIC (Takeuchi Information Criterion) TIC = -2ln(L) + 2tr(J^{-1}K), where J is Fisher information and K is outer product of scores

pub fn compare_models( &self, models: &[(String, f64, usize, usize)], criterion: InformationCriterion, ) -> Result<ModelComparisonResult>

Compare multiple models using specified criterion

pub fn model_averaged_prediction( &self, predictions: &[Array1<f64>], weights: &[f64], ) -> Result<Array1<f64>>

Calculate model-averaged prediction using IC weights

Trait Implementations

impl Default for InformationCriterionCalculator

fn default() -> Self

Returns the “default value” for a type.