Struct linregress::RegressionModel

source ·

pub struct RegressionModel { /* private fields */ }

Expand description

A fitted regression model.

Is the result of FormulaRegressionBuilder.fit().

Implementations§

source §

impl RegressionModel

source

pub fn regressor_names(&self) -> &[String]

The names of the regressor columns

source

pub fn p_values(&self) -> &[f64]

The two-tailed p-values for the t-statistics of the parameters

source

pub fn iter_p_value_pairs(&self) -> impl Iterator<Item = (&str, f64)> + '_

Iterates over pairs of regressor columns and their associated p-values

Note

This does not include the value for the intercept.

Usage

use linregress::{FormulaRegressionBuilder, RegressionDataBuilder};

let y = vec![1.,2. ,3. , 4.];
let x1 = vec![4., 3., 2., 1.];
let x2 = vec![1., 2., 3., 4.];
let data = vec![("Y", y), ("X1", x1), ("X2", x2)];
let data = RegressionDataBuilder::new().build_from(data)?;
let model = FormulaRegressionBuilder::new().data(&data).formula("Y ~ X1 + X2").fit()?;
let pairs: Vec<(&str, f64)> = model.iter_p_value_pairs().collect();
assert_eq!(pairs[0], ("X1", 1.7052707580549508e-28));
assert_eq!(pairs[1], ("X2", 2.522589878779506e-31));

source

pub fn residuals(&self) -> &[f64]

The residuals of the model

source

pub fn parameters(&self) -> &[f64]

The model’s intercept and slopes (also known as betas)

source

pub fn iter_parameter_pairs(&self) -> impl Iterator<Item = (&str, f64)> + '_

Iterates over pairs of regressor columns and their associated slope values

Note

This does not include the value for the intercept.

Usage

use linregress::{FormulaRegressionBuilder, RegressionDataBuilder};

let y = vec![1.,2. ,3. , 4.];
let x1 = vec![4., 3., 2., 1.];
let x2 = vec![1., 2., 3., 4.];
let data = vec![("Y", y), ("X1", x1), ("X2", x2)];
let data = RegressionDataBuilder::new().build_from(data)?;
let model = FormulaRegressionBuilder::new().data(&data).formula("Y ~ X1 + X2").fit()?;
let pairs: Vec<(&str, f64)> = model.iter_parameter_pairs().collect();
assert_eq!(pairs[0], ("X1", -0.03703703703703709));
assert_eq!(pairs[1], ("X2", 0.9629629629629626));

source

pub fn se(&self) -> &[f64]

The standard errors of the parameter estimates

source

pub fn iter_se_pairs(&self) -> impl Iterator<Item = (&str, f64)> + '_

Iterates over pairs of regressor columns and their associated standard errors

Note

This does not include the value for the intercept.

Usage

use linregress::{FormulaRegressionBuilder, RegressionDataBuilder};

let y = vec![1.,2. ,3. , 4.];
let x1 = vec![4., 3., 2., 1.];
let x2 = vec![1., 2., 3., 4.];
let data = vec![("Y", y), ("X1", x1), ("X2", x2)];
let data = RegressionDataBuilder::new().build_from(data)?;
let model = FormulaRegressionBuilder::new().data(&data).formula("Y ~ X1 + X2").fit()?;
let pairs: Vec<(&str, f64)> = model.iter_parameter_pairs().collect();
assert_eq!(pairs[0], ("X1", -0.03703703703703709));
assert_eq!(pairs[1], ("X2", 0.9629629629629626));

source

pub fn ssr(&self) -> f64

Sum of squared residuals

source

pub fn rsquared(&self) -> f64

R-squared of the model

source

pub fn rsquared_adj(&self) -> f64

Adjusted R-squared of the model

source

pub fn scale(&self) -> f64

A scale factor for the covariance matrix

Note that the square root of scale is often called the standard error of the regression.

source

pub fn predict<'a, I, S>(&self, new_data: I) -> Result<Vec<f64>, Error>where I: IntoIterator<Item = (S, Vec<f64>)>, S: Into<Cow<'a, str>>,

Evaluates the model on given new input data and returns the predicted values.

The new data is expected to have the same columns as the original data. See RegressionDataBuilder.build for details on the type of the new_data parameter.

Note

This function does no special handling of non real values (NaN or infinity or negative infinity). Such a value in new_data will result in a corresponding meaningless prediction.

Example

let y = vec![1., 2., 3., 4., 5.];
let x1 = vec![5., 4., 3., 2., 1.];
let x2 = vec![729.53, 439.0367, 42.054, 1., 0.];
let x3 = vec![258.589, 616.297, 215.061, 498.361, 0.];
let data = vec![("Y", y), ("X1", x1), ("X2", x2), ("X3", x3)];
let data = RegressionDataBuilder::new().build_from(data).unwrap();
let formula = "Y ~ X1 + X2 + X3";
let model = FormulaRegressionBuilder::new()
    .data(&data)
    .formula(formula)
    .fit()?;
let new_data = vec![
    ("X1", vec![2.5, 3.5]),
    ("X2", vec![2.0, 8.0]),
    ("X3", vec![2.0, 1.0]),
];
let prediction: Vec<f64> = model.predict(new_data)?;
assert_slices_almost_eq!(&prediction, &[3.500000000000028, 2.5000000000000644]);