Struct linregress::FormulaRegressionBuilder

pub struct FormulaRegressionBuilder<'a> { /* private fields */ }

A builder to create and fit a linear regression model.

Given a dataset and a set of columns to use this builder will produce an ordinary least squared linear regression model.

See formula and data for details on how to configure this builder.

The pseudo inverse method is used to fit the model.

Usage

use linregress::{FormulaRegressionBuilder, RegressionDataBuilder, assert_almost_eq};

let y = vec![1., 2. ,3., 4.];
let x = vec![4., 3., 2., 1.];
let data = vec![("Y", y), ("X", x)];
let data = RegressionDataBuilder::new().build_from(data)?;
let model = FormulaRegressionBuilder::new().data(&data).formula("Y ~ X").fit()?;
// Alternatively
let model = FormulaRegressionBuilder::new().data(&data).data_columns("Y", ["X"]).fit()?;
let params = model.parameters();
assert_almost_eq!(params[0], 4.999999999999998);
assert_almost_eq!(params[1], -0.9999999999999989);
assert_eq!(model.regressor_names()[0], "X");

Implementations

impl<'a> FormulaRegressionBuilder<'a>

pub fn new() -> Self

Create as new FormulaRegressionBuilder with no data or formula set.

pub fn data(self, data: &'a RegressionData<'a>) -> Self

Set the data to be used for the regression.

The data has to be given as a reference to a RegressionData struct. See RegressionDataBuilder for details.

pub fn formula<T: Into<Cow<'a, str>>>(self, formula: T) -> Self

Set the formula to use for the regression.

The expected format is <regressand> ~ <regressor 1> + <regressor 2>.

E.g. for a regressand named Y and three regressors named A, B and C the correct format would be Y ~ A + B + C.

Note that there is currently no special support for categorical variables. So if you have a categorical variable with more than two distinct values or values that are not 0 and 1 you will need to perform “dummy coding” yourself.

Alternatively you can use data_columns.

pub fn data_columns<I, S1, S2>(self, regressand: S1, regressors: I) -> Selfwhere I: IntoIterator<Item = S2>, S1: Into<Cow<'a, str>>, S2: Into<Cow<'a, str>>,

Set the columns to be used as regressand and regressors for the regression.

Note that there is currently no special support for categorical variables. So if you have a categorical variable with more than two distinct values or values that are not 0 and 1 you will need to perform “dummy coding” yourself.

Alternatively you can use formula.

pub fn fit(self) -> Result<RegressionModel, Error>

Fits the model and returns a RegressionModel if successful. You need to set the data with data and a formula with formula before you can use it.

pub fn fit_without_statistics(self) -> Result<Vec<f64>, Error>

Like fit but does not perfom any statistics on the resulting model. Returns a Vec containing the model parameters (in the order intercept, column 1, column 2, …) if successfull.

This is usefull if you do not care about the statistics or the model and data you want to fit result in too few residual degrees of freedom to perform statistics.

Trait Implementations

impl<'a> Clone for FormulaRegressionBuilder<'a>

fn clone(&self) -> FormulaRegressionBuilder<'a>

Returns a copy of the value.

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

Performs copy-assignment from source.

impl<'a> Debug for FormulaRegressionBuilder<'a>

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

Formats the value using the given formatter.

impl<'a> Default for FormulaRegressionBuilder<'a>

fn default() -> Self

Returns the “default value” for a type.