Struct LogisticFit 

pub struct LogisticFit { /* private fields */ }

A fitted binary logistic-regression model.

Fit by iteratively reweighted least squares (equivalently, Fisher scoring / Newton–Raphson on the log-likelihood). At convergence the score equations Xᵀ(y − p) = 0 hold, and the inverse Fisher information (XᵀWX)⁻¹ — with W = diag(pᵢ(1−pᵢ)) — gives the coefficient covariance the Wald statistics are built from.

See the module docs for the response convention and the separation caveat.

Implementations

impl LogisticFit

pub fn new(x: Array2<f64>, y: Array1<f64>) -> Result<Self>

Fit logistic regression of binary y on X (default: up to 100 IRLS iterations, tolerance 1e-10 on the coefficient step).

Errors

• RegressionError::EmptyInput / RegressionError::ShapeMismatch.
• RegressionError::InvalidResponse if y is not 0/1 or is all one class.
• RegressionError::RankDeficient if the (weighted) design is singular.
• RegressionError::NotConverged if IRLS fails to converge (typically separation).

pub fn with_options( x: Array2<f64>, y: Array1<f64>, max_iter: usize, tol: f64, ) -> Result<Self>

Like LogisticFit::new with an explicit iteration cap and tolerance.

pub fn n_observations(&self) -> usize

Number of observations.

pub fn n_parameters(&self) -> usize

Number of coefficients (design columns, intercept included).

pub fn has_intercept(&self) -> bool

Whether an intercept (constant) column is present.

pub fn iterations(&self) -> usize

IRLS iterations taken to converge.

pub fn design_matrix(&self) -> ArrayView2<'_, f64>

The design matrix as fitted.

pub fn response(&self) -> ArrayView1<'_, f64>

The binary response.

pub fn coefficients(&self) -> ArrayView1<'_, f64>

Estimated coefficients (log-odds scale), aligned to the design columns.

pub fn fitted_probabilities(&self) -> ArrayView1<'_, f64>

Fitted probabilities pᵢ = P(yᵢ = 1).

pub fn weights(&self) -> ArrayView1<'_, f64>

IRLS weights wᵢ = pᵢ(1 − pᵢ) at the MLE.

pub fn covariance(&self) -> ArrayView2<'_, f64>

Coefficient covariance matrix (XᵀWX)⁻¹.

pub fn log_likelihood(&self) -> f64

Maximized log-likelihood.

pub fn coefficient_standard_errors(&self) -> Array1<f64>

Coefficient standard errors √diag((XᵀWX)⁻¹).

pub fn z_values(&self) -> Array1<f64>

Wald z-statistics βⱼ / seⱼ.

pub fn p_values(&self) -> Array1<f64>

Two-sided Wald p-values from the standard normal.

pub fn predict_proba(&self, x: ArrayView2<'_, f64>) -> Array1<f64>

Predicted probabilities for a new design matrix x (same column layout as the training design).

impl LogisticFit

pub fn goodness_of_fit(&self) -> GoodnessOfFit

Overall goodness-of-fit summary: null/residual deviance, McFadden’s pseudo-R², and AIC/BIC.

The residual deviance is −2ℓ (the saturated-model log-likelihood is zero for ungrouped binary data), and equals the sum of squared deviance residuals. McFadden’s pseudo-R² compares the fitted log-likelihood to the intercept-only model; it is bounded in [0, 1) but runs lower than an OLS R² for comparable fits, so judge it on that scale.

pub fn hosmer_lemeshow(&self, groups: usize) -> HosmerLemeshow

Hosmer–Lemeshow goodness-of-fit test with groups risk deciles (groups = 10 is the customary choice).

Observations are ordered by fitted probability and split into groups near-equal bins; the statistic compares observed and expected event counts per bin,

Ĥ = Σ_g (O_g − E_g)² / (n_g · π̄_g · (1 − π̄_g)),

which is asymptotically χ²(groups − 2). Unlike most tests here, a small p-value means the model fits poorly. Bins whose mean probability is exactly 0 or 1 contribute nothing (their variance term is undefined).

The test is only meaningful when groups ≥ 3 and there are enough observations to populate the bins; with fewer than groups + 1 observations the statistic is returned as NaN.

Trait Implementations

impl Clone for LogisticFit

fn clone(&self) -> LogisticFit

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for LogisticFit

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

Formats the value using the given formatter.