Struct NormalLikelihood 

pub struct NormalLikelihood {
    pub mean_parameterized: bool,
    pub variance_parameterized: bool,
    pub function_name: String,
    pub description: String,
    pub formula: String,
}

Likelihood for data under a normal model.

Fields

mean_parameterized: bool

Whether the mean is parameterized.

variance_parameterized: bool

Whether the variance is parameterized.

function_name: String

Unique name identifying a likelihood function.

description: String

Free-text description.

formula: String

Mathematical formula.

Implementations

impl NormalLikelihood

pub fn fit(&self, data: &[f64]) -> Result<MleFit>

Closed-form maximum-likelihood fit of θ = [μ, σ] to data.

The Gaussian MLE is analytic: μ̂ is the sample mean and σ̂ the biased (population) standard deviation √(Σ(xᵢ − μ̂)²/n) — matching numpy.std(data, ddof=0). The result is exact, so the returned MleFit reports converged() == true and iterations() == 0.

Arguments

• data — the observed sample; must contain at least two points with non-zero spread.

Returns

An MleFit whose params() are [μ̂, σ̂], carrying the attained log-likelihood and the AIC/BIC (k = 2, n = data.len()).

Errors

• Error::InsufficientData if data has fewer than two points (with a single point σ̂ = 0, which the density cannot represent).
• Error::DegenerateInput if every observation is identical, so the estimated σ̂ is zero and the log-likelihood is undefined.

Examples

use stats_claw::likelihood::NormalLikelihood;

let fit = NormalLikelihood::default().fit(&[1.0, 2.0, 3.0])?;
assert!((fit.params()[0] - 2.0).abs() < 1e-12, "mu_hat was {}", fit.params()[0]);
assert!(fit.converged());

Trait Implementations

impl Clone for NormalLikelihood

fn clone(&self) -> NormalLikelihood

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for NormalLikelihood

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

Formats the value using the given formatter.

impl Default for NormalLikelihood

fn default() -> NormalLikelihood

Returns the “default value” for a type.

impl LogLikelihood for NormalLikelihood

fn n_params(&self) -> usize

Returns 2 — the model’s free parameters are μ (mean) and σ (standard deviation), in that order.

fn log_likelihood(&self, params: &[f64], data: &[f64]) -> f64

Evaluates the Gaussian log-likelihood ℓ = −n/2·ln(2π) − n·ln σ − Σ(xᵢ − μ)²/(2σ²) with params = [μ, σ].

Returns f64::NEG_INFINITY for any θ outside the valid domain: a non-positive or non-finite σ, or a non-finite μ. Per the LogLikelihood contract, any non-finite observation (NaN or ±∞) also yields f64::NEG_INFINITY rather than letting (xᵢ − μ)² propagate a NaN (or an incidental −∞ for +∞). An empty data yields 0.0 (the empty product’s log-likelihood); callers wanting an error on empty input use NormalLikelihood::fit.