wls-rust

weighted linear regression in pure Rust w/o any 3d party dependencies or frameworks.

How-to

Fit linear regression using weights

fn main() {
    let x_points = [1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0];
    let y_points = [1.0, 3.0, 4.0, 5.0, 2.0, 3.0, 4.0];
    let weights = [10.0, 1.0, 3.0, 8.0, 14.0, 21.0, 13.0];

    let wls = Wls::new(&x_points, &y_points, &weights);
    let point = wls.fit_linear_regression().unwrap();

    assert_almost_equal(1.410964913, point.get_intercept(), 1.0e-6);
    assert_almost_equal(0.321271930, point.get_slope(), 1.0e-6);
}

Fit linear regression using stable weights

fn main() {
    let x_points = vec![1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0];
    let y_points = vec![1.0, 3.0, 4.0, 5.0, 2.0, 3.0, 4.0];
    let stable_weights = vec![1.0; x_points.len()];

    let wls = Wls::new(&x_points, &y_points, &stable_weights);
    let point = wls.fit_linear_regression().unwrap();

    assert_almost_equal(2.14285714, point.get_intercept(), 1.0e-6);
    assert_eq!(0.25, point.get_slope());
}

Description

WLS is based on the OLS method and help solve problems of model inadequacy or violations of the basic regression assumptions.

Estimating a linear regression with WLS is useful, but can appear to be daunting w/o special stats packages, such as Python statsmodels or Pandas.

References

• Wikipedia: Weighted least squares
• Introduction to Linear Regression Analysis, 5th edition
• Least Squares Regression Analysis in Terms of Linear Algebra