pub fn simple_kde(
values: &[f64],
bandwidth: f64,
samples: usize,
) -> Vec<(f64, f64)>Expand description
Gaussian kernel density estimate. Extends the evaluation range by 3*bandwidth on each side so Gaussian tails taper smoothly rather than terminating sharply at the data extremes.
Uses a truncated kernel: for each evaluation point only the sorted values within 4*bandwidth contribute (Gaussian contribution beyond that is < 0.003%). This gives O(window × samples) instead of O(n × samples).