Struct compute::predict::GLM

pub struct GLM {
    pub family: ExponentialFamily,
    pub alpha: f64,
    pub tolerance: f64,
    pub weights: Option<Vec<f64>>,
    // some fields omitted
}

Implements a generalized linear model.

Fields

family: ExponentialFamily

alpha: f64

tolerance: f64

weights: Option<Vec<f64>>

Implementations

impl GLM

pub fn new(family: ExponentialFamily) -> Self

Create a new general linear model with the given exponential family. alpha sets the L2 (ridge regression) regularization strength, and tolerance sets the convergence tolerance.

pub fn set_penalty(&mut self, alpha: f64) -> &mut Self

Set the L2 (ridge regression) regularization strength.

pub fn set_tolerance(&mut self, tolerance: f64) -> &mut Self

Set the convergence tolerance.

pub fn set_coef(&mut self, coefs: &[f64]) -> &mut Self

Set the model coefficients.

pub fn set_weights(&mut self, weights: &[f64]) -> &mut Self

Set the sample weights (usually measurement errors).

pub fn set_offset(&mut self, offset: &[f64]) -> &mut Self

Set the offsets (usually used in Poisson regression models).

pub fn fit(&mut self, x: &[f64], y: &[f64], max_iter: usize) -> Result<(), &str>

Fit the GLM using the scoring algorithm, which gives the maximumum likelihood estimate. It performs a maximum of max_iter iterations. Note that x must be a design matrix (i.e., the first column must contain all 1’s).

pub fn coef(&self) -> Result<&[f64], &str>

Return the maximum likelihood estimates for the parameters.

pub fn deviance(&self) -> Result<f64, &str>

Return the deviance of the model.

pub fn aic(&self) -> Result<f64, &str>

Calculates the Akaike information criterion for the model.

pub fn bic(&self) -> Result<f64, &str>

Calculates the Bayesian information criterion for the model.

pub fn dispersion(&self) -> Result<f64, &str>

Calculates the dispersion of the model.

pub fn coef_covariance_matrix(&self) -> Result<Vec<f64>, &str>

Returns the fitted covariance for the estimated parameters.

pub fn coef_standard_error(&self) -> Result<Vec<f64>, &str>

Returns the estimated standard errors on the estimated parameters. This is equivalent to the square root of the diagonals of the covariance matrix.

pub fn predict(&self, x: &[f64]) -> Result<Vector, &str>

Use the fitted model to make predictions on some new data.

pub fn score(&self, x: &[f64], y: &[f64]) -> f64

Make some new predictions and calculate the score of those predictions based on known responses.

Trait Implementations

impl Clone for GLM

fn clone(&self) -> GLM

Returns a copy of the value.

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

Performs copy-assignment from source.

impl Debug for GLM

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

Formats the value using the given formatter.