1 use crate::stats::univariate::kde::kernel::Gaussian;
2 use crate::stats::univariate::kde::{Bandwidth, Kde};
3 use crate::stats::univariate::Sample;
4
sweep( sample: &Sample<f64>, npoints: usize, range: Option<(f64, f64)>, ) -> (Box<[f64]>, Box<[f64]>)5 pub fn sweep(
6 sample: &Sample<f64>,
7 npoints: usize,
8 range: Option<(f64, f64)>,
9 ) -> (Box<[f64]>, Box<[f64]>) {
10 let (xs, ys, _) = sweep_and_estimate(sample, npoints, range, sample[0]);
11 (xs, ys)
12 }
13
sweep_and_estimate( sample: &Sample<f64>, npoints: usize, range: Option<(f64, f64)>, point_to_estimate: f64, ) -> (Box<[f64]>, Box<[f64]>, f64)14 pub fn sweep_and_estimate(
15 sample: &Sample<f64>,
16 npoints: usize,
17 range: Option<(f64, f64)>,
18 point_to_estimate: f64,
19 ) -> (Box<[f64]>, Box<[f64]>, f64) {
20 let x_min = sample.min();
21 let x_max = sample.max();
22
23 let kde = Kde::new(sample, Gaussian, Bandwidth::Silverman);
24 let h = kde.bandwidth();
25
26 let (start, end) = match range {
27 Some((start, end)) => (start, end),
28 None => (x_min - 3. * h, x_max + 3. * h),
29 };
30
31 let mut xs: Vec<f64> = Vec::with_capacity(npoints);
32 let step_size = (end - start) / (npoints - 1) as f64;
33 for n in 0..npoints {
34 xs.push(start + (step_size * n as f64));
35 }
36
37 let ys = kde.map(&xs);
38 let point_estimate = kde.estimate(point_to_estimate);
39
40 (xs.into_boxed_slice(), ys, point_estimate)
41 }
42