Function chi2 

pub fn chi2<F: Float + Send + Sync + 'static>(
    x: &Array2<F>,
    y: &Array1<usize>,
) -> Result<(Array1<F>, Array1<F>), FerroError>

Compute chi-squared statistics between each non-negative feature and the class labels.

For each feature the observed and expected frequencies per class are computed, then:

chi2 = sum_class (observed - expected)^2 / expected

where observed is the sum of feature values for samples of that class, and expected is the expected sum under the null hypothesis (proportional to the class frequency and the overall feature sum).

Returns

(chi2_statistics, p_values) — two Array1<F> of length n_features.

Errors

• FerroError::InsufficientSamples if x has zero rows.
• FerroError::ShapeMismatch if x.nrows() != y.len().
• FerroError::InvalidParameter if any feature value is negative.

Examples

use ferrolearn_preprocess::feature_scoring::chi2;
use ndarray::{array, Array1};

let x = array![[1.0_f64, 0.0], [0.0, 1.0], [1.0, 0.0], [0.0, 1.0]];
let y: Array1<usize> = array![0, 1, 0, 1];
let (chi2_stats, _p_vals) = chi2(&x, &y).unwrap();
assert_eq!(chi2_stats.len(), 2);