Kde with auto kernel width

This commit is contained in:
Ozzie Gooen 2020-04-03 20:41:37 +01:00
parent 1b541ec038
commit 7ab81249bf
6 changed files with 37 additions and 35 deletions

View File

@ -4,10 +4,10 @@ open Expect;
describe("Bandwidth", () => {
test("nrd0()", () => {
let data = [|1., 4., 3., 2.|];
expect(Science.nrd0(data)) |> toEqual(0.7635139420854616);
expect(Bandwidth.nrd0(data)) |> toEqual(0.7625801874014622);
});
test("nrd()", () => {
let data = [|1., 4., 3., 2.|];
expect(Science.nrd(data)) |> toEqual(0.899249754011766);
expect(Bandwidth.nrd(data)) |> toEqual(0.8981499984950554);
});
});

View File

@ -60,7 +60,6 @@
"react-use": "^13.27.0",
"reason-react": ">=0.7.0",
"reschema": "1.3.0",
"science": "^1.9.3",
"tailwindcss": "1.2.0"
},
"alias": {

View File

@ -89,11 +89,10 @@ module T = {
};
let kde = (~samples, ~outputXYPoints) => {
let width = Science.nrd0(samples);
let width = Bandwidth.nrd0(samples);
let xyPointRange = E.A.Sorted.range(samples) |> E.O.default(0.0);
let xyPointWidth = xyPointRange /. float_of_int(outputXYPoints);
let kernelWidth = int_of_float(Jstat.max([|(width /. xyPointWidth), 1.0 |]));
Js.log4(samples, width, xyPointWidth, kernelWidth);
KDE.normalSampling(samples, outputXYPoints, kernelWidth);
};

View File

@ -97,6 +97,8 @@ type binomial = {
[@bs.module "jstat"] external deviation: array(float) => float = "deviation";
[@bs.module "jstat"] external stdev: array(float) => float = "stdev";
[@bs.module "jstat"]
external quantiles: (array(float), array(float)) => float = "quantiles";
external quartiles: (array(float)) => array(float) = "quartiles";
[@bs.module "jstat"]
external percentile: (array(float), array(float)) => float = "percentile";
external quantiles: (array(float), array(float)) => array(float) = "quantiles";
[@bs.module "jstat"]
external percentile: (array(float), float, bool) => float = "percentile";

30
src/utility/Bandwidth.re Normal file
View File

@ -0,0 +1,30 @@
//The math here was taken from https://github.com/jasondavies/science.js/blob/master/src/stats/bandwidth.js
let len = x => E.A.length(x) |> float_of_int;
let iqr = x => {
Jstat.percentile(x, 0.75, true) -. Jstat.percentile(x, 0.25, true);
};
// Silverman, B. W. (1986) Density Estimation. London: Chapman and Hall.
let nrd0 = x => {
let hi = Js_math.sqrt(Jstat.variance(x));
let lo = Js_math.minMany_float([|hi, iqr(x) /. 1.34|]);
let e = Js_math.abs_float(x[1]);
let lo' =
switch (lo, hi, e) {
| (lo, _, _) when !Js.Float.isNaN(lo) => lo
| (_, hi, _) when !Js.Float.isNaN(hi) => hi
| (_, _, e) when !Js.Float.isNaN(e) => e
| _ => 1.0
};
0.9 *. lo' *. Js.Math.pow_float(~base=len(x), ~exp=-0.2);
};
// Scott, D. W. (1992) Multivariate Density Estimation: Theory, Practice, and Visualization. Wiley.
let nrd = x => {
let h = iqr(x) /. 1.34;
1.06
*. Js.Math.min_float(Js.Math.sqrt(Jstat.variance(x)), h)
*. Js.Math.pow_float(~base=len(x), ~exp=(-1.0) /. 5.0);
};

View File

@ -1,28 +0,0 @@
[@bs.val] [@bs.module "science"] [@bs.scope "stats"]
external variance: array(float) => float = "variance";
[@bs.val] [@bs.module "science"] [@bs.scope "stats"]
external iqr: array(float) => float = "iqr";
let len = x => E.A.length(x) |> float_of_int;
let nrd0 = x => {
let hi = Js_math.sqrt(variance(x));
let lo = Js_math.minMany_float([|hi, iqr(x) /. 1.34|]);
let e = Js_math.abs_float(x[1]);
let lo' =
switch (lo, hi, e) {
| (lo, hi, e) when !Js.Float.isNaN(lo) => lo
| (lo, hi, e) when !Js.Float.isNaN(hi) => hi
| (lo, hi, e) when !Js.Float.isNaN(e) => e
| _ => 1.0
};
0.9 *. lo' *. Js_math.pow_float(len(x), -0.2);
};
let nrd = x => {
let h = iqr(x) /. 1.34;
1.06
*. Js_math.min_float(Js_math.sqrt(variance(x)), h)
*. Js_math.pow_float(len(x), (-1.0) /. 5.0);
};