Kendall's tau rank correlation. Initially the library was based on Apache Commons Math library with some additions taken from scipy and R cor.test function.

Example usage:

let (tau_b, significance) = kendalls::tau_b(&[1, 2, 3], &[3, 4, 5]).unwrap();
assert_eq!(tau_b, 1.0);
assert_eq!(significance, 1.5666989036012806);

If you want to compute correlation, let's say, for f64 type, then you will have to provide either a custom comparator function or declare Ord trait for your custom floating point numbers type (see float crate).

use std::cmp::Ordering;

let (tau_b, _significance) = kendalls::tau_b_with_comparator(
    &[1.0, 2.0],
    &[3.0, 4.0],
    |a: &f64, b: &f64| a.partial_cmp(&b).unwrap_or(Ordering::Greater),
).unwrap();
assert_eq!(tau_b, 1.0);

The function will return an error if you pass empty arrays into it or x and y arrays' dimensions are not equal.

kendalls

Kendall's rank correlation coefficient

Usage

Add this to your Cargo.toml:

[dependencies]
kendalls = "1.0.0"

and this to your crate root:

extern crate kendalls;

Example:

fn main() -> Result<(), kendalls::Error> {
    let (tau_b, significance) = kendalls::tau_b(&[1, 2, 3], &[3, 4, 5])?;
    assert_eq!(tau_b, 1.0);
    assert_eq!(significance, 1.5666989036012806);

    Ok(())
}