Function fpc_shap_values 

pub fn fpc_shap_values(
    fit: &FregreLmResult,
    data: &FdMatrix,
    scalar_covariates: Option<&FdMatrix>,
) -> Result<FpcShapValues, FdarError>

Exact SHAP values for a linear functional regression model.

For linear models, SHAP values are exact: values[(i,k)] = coef[1+k] * (score_i_k - mean_k). The efficiency property holds: base_value + sum_k values[(i,k)] ~ fitted_values[i] (with scalar covariate effects absorbed into the base value).

Errors

Returns FdarError::InvalidDimension if data has zero rows or its column count does not match fit.fpca.mean. Returns FdarError::InvalidParameter if fit.ncomp is zero.

Examples

use fdars_core::matrix::FdMatrix;
use fdars_core::scalar_on_function::fregre_lm;
use fdars_core::explain::fpc_shap_values;

let (n, m) = (20, 30);
let data = FdMatrix::from_column_major(
    (0..n * m).map(|k| {
        let i = (k % n) as f64;
        let j = (k / n) as f64;
        ((i + 1.0) * j * 0.2).sin()
    }).collect(),
    n, m,
).unwrap();
let y: Vec<f64> = (0..n).map(|i| (i as f64 * 0.5).sin()).collect();
let fit = fregre_lm(&data, &y, None, 3).unwrap();
let shap = fpc_shap_values(&fit, &data, None).unwrap();
assert_eq!(shap.values.shape(), (20, 3));