Socket
Socket
Sign inDemoInstall

vega-statistics

Package Overview
Dependencies
Maintainers
1
Versions
32
Alerts
File Explorer

Advanced tools

License
Socket logo

Install Socket

Detect and block malicious and high-risk dependencies

Install

vega-statistics - npm Package Compare versions

Comparing version 1.8.0 to 1.8.1

rollup.config.mjs

579

build/vega-statistics.js

@@ -16,6 +16,4 @@ (function (global, factory) {

let index = -1;
for (let value of values) {
value = valueof(value, ++index, values);
if (value != null && value !== '' && (value = +value) >= value) {

@@ -44,3 +42,2 @@ yield value;

let index = -1;
for (let value of values) {

@@ -59,3 +56,2 @@ if ((value = valueof(value, ++index, values)) != null && (value = +value) >= value) {

let sum = 0;
if (valueof === undefined) {

@@ -71,3 +67,2 @@ for (let value of values) {

let index = -1;
for (let value of values) {

@@ -81,3 +76,2 @@ if ((value = valueof(value, ++index, values)) != null && (value = +value) >= value) {

}
if (count > 1) return sum / (count - 1);

@@ -106,3 +100,2 @@ }

let max;
if (valueof === undefined) {

@@ -116,3 +109,2 @@ for (const value of values) {

let index = -1;
for (let value of values) {

@@ -124,3 +116,2 @@ if ((value = valueof(value, ++index, values)) != null && (max < value || max === undefined && value >= value)) {

}
return max;

@@ -131,3 +122,2 @@ }

let min;
if (valueof === undefined) {

@@ -141,3 +131,2 @@ for (const value of values) {

let index = -1;
for (let value of values) {

@@ -149,11 +138,13 @@ if ((value = valueof(value, ++index, values)) != null && (min > value || min === undefined && value >= value)) {

}
return min;
}
// Based on https://github.com/mourner/quickselect
// ISC license, Copyright 2018 Vladimir Agafonkin.
function quickselect(array, k, left = 0, right = array.length - 1, compare) {
function quickselect(array, k, left = 0, right = Infinity, compare) {
k = Math.floor(k);
left = Math.floor(Math.max(0, left));
right = Math.floor(Math.min(array.length - 1, right));
if (!(left <= k && k <= right)) return array;
compare = compare === undefined ? ascendingDefined : compareDefined(compare);
while (right > left) {

@@ -170,3 +161,2 @@ if (right - left > 600) {

}
const t = array[k];

@@ -177,11 +167,7 @@ let i = left;

if (compare(array[right], t) > 0) swap(array, left, right);
while (i < j) {
swap(array, i, j), ++i, --j;
while (compare(array[i], t) < 0) ++i;
while (compare(array[j], t) > 0) --j;
}
if (compare(array[left], t) === 0) swap(array, left, j);else ++j, swap(array, j, right);

@@ -191,6 +177,4 @@ if (j <= k) left = j + 1;

}
return array;
}
function swap(array, i, j) {

@@ -204,21 +188,21 @@ const t = array[i];

values = Float64Array.from(numbers(values, valueof));
if (!(n = values.length)) return;
if ((p = +p) <= 0 || n < 2) return min(values);
if (!(n = values.length) || isNaN(p = +p)) return;
if (p <= 0 || n < 2) return min(values);
if (p >= 1) return max(values);
var n,
i = (n - 1) * p,
i0 = Math.floor(i),
value0 = max(quickselect(values, i0).subarray(0, i0 + 1)),
value1 = min(values.subarray(i0 + 1));
i = (n - 1) * p,
i0 = Math.floor(i),
value0 = max(quickselect(values, i0).subarray(0, i0 + 1)),
value1 = min(values.subarray(i0 + 1));
return value0 + (value1 - value0) * (i - i0);
}
function quantileSorted(values, p, valueof = number) {
if (!(n = values.length)) return;
if ((p = +p) <= 0 || n < 2) return +valueof(values[0], 0, values);
if (!(n = values.length) || isNaN(p = +p)) return;
if (p <= 0 || n < 2) return +valueof(values[0], 0, values);
if (p >= 1) return +valueof(values[n - 1], n - 1, values);
var n,
i = (n - 1) * p,
i0 = Math.floor(i),
value0 = +valueof(values[i0], i0, values),
value1 = +valueof(values[i0 + 1], i0 + 1, values);
i = (n - 1) * p,
i0 = Math.floor(i),
value0 = +valueof(values[i0], i0, values),
value1 = +valueof(values[i0 + 1], i0 + 1, values);
return value0 + (value1 - value0) * (i - i0);

@@ -232,5 +216,6 @@ }

function quantiles (array, p, f) {
const values = Float64Array.from(numbers$1(array, f)); // don't depend on return value from typed array sort call
const values = Float64Array.from(numbers$1(array, f));
// don't depend on return value from typed array sort call
// protects against undefined sort results in Safari (vega/vega-lite#4964)
values.sort(ascending);

@@ -244,10 +229,10 @@ return p.map(_ => quantileSorted(values, _));

// Scott, D. W. (1992) Multivariate Density Estimation:
// Theory, Practice, and Visualization. Wiley.
function estimateBandwidth (array, f) {
const n = array.length,
d = deviation(array, f),
q = quartiles(array, f),
h = (q[2] - q[0]) / 1.34,
v = Math.min(d, h) || d || Math.abs(q[0]) || 1;
d = deviation(array, f),
q = quartiles(array, f),
h = (q[2] - q[0]) / 1.34,
v = Math.min(d, h) || d || Math.abs(q[0]) || 1;
return 1.06 * v * Math.pow(n, -0.2);

@@ -259,15 +244,14 @@ }

const maxb = _.maxbins || 20,
base = _.base || 10,
logb = Math.log(base),
div = _.divide || [5, 2];
base = _.base || 10,
logb = Math.log(base),
div = _.divide || [5, 2];
let min = _.extent[0],
max = _.extent[1],
step,
level,
minstep,
v,
i,
n;
max = _.extent[1],
step,
level,
minstep,
v,
i,
n;
const span = _.span || max - min || Math.abs(min) || 1;
if (_.step) {

@@ -279,5 +263,3 @@ // if step size is explicitly given, use that

v = span / maxb;
for (i = 0, n = _.steps.length; i < n && _.steps[i] < v; ++i);
step = _.steps[Math.max(0, i - 1)];

@@ -288,9 +270,10 @@ } else {

minstep = _.minstep || 0;
step = Math.max(minstep, Math.pow(base, Math.round(Math.log(span) / logb) - level)); // increase step size if too many bins
step = Math.max(minstep, Math.pow(base, Math.round(Math.log(span) / logb) - level));
// increase step size if too many bins
while (Math.ceil(span / step) > maxb) {
step *= base;
} // decrease step size if allowed
}
// decrease step size if allowed
for (i = 0, n = div.length; i < n; ++i) {

@@ -300,9 +283,8 @@ v = step / div[i];

}
} // update precision, min and max
}
// update precision, min and max
v = Math.log(step);
const precision = v >= 0 ? 0 : ~~(-v / logb) + 1,
eps = Math.pow(base, -precision - 1);
eps = Math.pow(base, -precision - 1);
if (_.nice || _.nice === undefined) {

@@ -313,3 +295,2 @@ v = Math.floor(min / step + eps) * step;

}
return {

@@ -330,6 +311,5 @@ start: min,

const values = Float64Array.from(numbers$1(array, f)),
n = values.length,
m = samples;
n = values.length,
m = samples;
let a, i, j, mu;
for (j = 0, mu = Array(m); j < m; ++j) {

@@ -339,6 +319,4 @@ for (a = 0, i = 0; i < n; ++i) {

}
mu[j] = a / n;
}
mu.sort(ascending);

@@ -353,66 +331,54 @@ return [quantile(mu, alpha / 2), quantile(mu, 1 - alpha / 2)];

f = f || (_ => _);
const n = array.length,
v = new Float64Array(n);
v = new Float64Array(n);
let i = 0,
j = 1,
a = f(array[0]),
b = a,
w = a + step,
x;
j = 1,
a = f(array[0]),
b = a,
w = a + step,
x;
for (; j < n; ++j) {
x = f(array[j]);
if (x >= w) {
b = (a + b) / 2;
for (; i < j; ++i) v[i] = b;
w = x + step;
a = x;
}
b = x;
}
b = (a + b) / 2;
for (; i < j; ++i) v[i] = b;
return smooth ? smoothing(v, step + step / 4) : v;
}
return smooth ? smoothing(v, step + step / 4) : v;
} // perform smoothing to reduce variance
// perform smoothing to reduce variance
// swap points between "adjacent" stacks
// Wilkinson defines adjacent as within step/4 units
function smoothing(v, thresh) {
const n = v.length;
let a = 0,
b = 1,
c,
d; // get left stack
b = 1,
c,
d;
// get left stack
while (v[a] === v[b]) ++b;
while (b < n) {
// get right stack
c = b + 1;
while (v[b] === v[c]) ++c;
while (v[b] === v[c]) ++c; // are stacks adjacent?
// are stacks adjacent?
// if so, compare sizes and swap as needed
if (v[b] - v[b - 1] < thresh) {
d = b + (a + c - b - b >> 1);
while (d < b) v[d++] = v[b];
while (d > b) v[d--] = v[a];
} // update left stack indices
}
// update left stack indices
a = b;
b = c;
}
return v;

@@ -435,3 +401,2 @@ }

}
let a, b, d;

@@ -448,3 +413,2 @@ const dist = {

},
max(_) {

@@ -459,11 +423,8 @@ if (arguments.length) {

},
sample() {
return a + Math.floor(d * exports.random());
},
pdf(x) {
return x === Math.floor(x) && x >= a && x < b ? 1 / d : 0;
},
cdf(x) {

@@ -473,7 +434,5 @@ const v = Math.floor(x);

},
icdf(p) {
return p >= 0 && p <= 1 ? a - 1 + Math.floor(p * d) : NaN;
}
};

@@ -491,6 +450,5 @@ return dist.min(min).max(max);

let x = 0,
y = 0,
rds,
c;
y = 0,
rds,
c;
if (nextSample === nextSample) {

@@ -505,9 +463,6 @@ x = nextSample;

} while (rds === 0 || rds > 1);
c = Math.sqrt(-2 * Math.log(rds) / rds); // Box-Muller transform
x *= c;
nextSample = y * c;
}
return mean + x * stdev;

@@ -519,5 +474,6 @@ }

return Math.exp(-0.5 * z * z) / (stdev * SQRT2PI);
} // Approximation from West (2009)
}
// Approximation from West (2009)
// Better Approximations to Cumulative Normal Functions
function cumulativeNormal(value, mean, stdev) {

@@ -527,5 +483,4 @@ mean = mean || 0;

const z = (value - mean) / stdev,
Z = Math.abs(z);
Z = Math.abs(z);
let cd;
if (Z > 37) {

@@ -536,3 +491,2 @@ cd = 0;

let sum;
if (Z < 7.07106781186547) {

@@ -563,13 +517,14 @@ sum = 3.52624965998911e-02 * Z + 0.700383064443688;

}
return z > 0 ? 1 - cd : cd;
} // Approximation of Probit function using inverse error function.
}
// Approximation of Probit function using inverse error function.
function quantileNormal(p, mean, stdev) {
if (p < 0 || p > 1) return NaN;
return (mean || 0) + (stdev == null ? 1 : stdev) * SQRT2 * erfinv(2 * p - 1);
} // Approximate inverse error function. Implementation from "Approximating
}
// Approximate inverse error function. Implementation from "Approximating
// the erfinv function" by Mike Giles, GPU Computing Gems, volume 2, 2010.
// Ported from Apache Commons Math, http://www.apache.org/licenses/LICENSE-2.0
function erfinv(x) {

@@ -581,4 +536,3 @@ // beware that the logarithm argument must be

let w = -Math.log((1 - x) * (1 + x)),
p;
p;
if (w < 6.25) {

@@ -652,6 +606,4 @@ w -= 3.125;

}
return p * x;
}
function gaussian (mean, stdev) {

@@ -668,3 +620,2 @@ let mu, sigma;

},
stdev(_) {

@@ -678,3 +629,2 @@ if (arguments.length) {

},
sample: () => sampleNormal(mu, sigma),

@@ -701,3 +651,2 @@ pdf: value => densityNormal(value, mu, sigma),

},
bandwidth(_) {

@@ -709,33 +658,24 @@ if (!arguments.length) return bandwidth;

},
sample() {
return support[~~(exports.random() * n)] + bandwidth * kernel.sample();
},
pdf(x) {
let y = 0,
i = 0;
i = 0;
for (; i < n; ++i) {
y += kernel.pdf((x - support[i]) / bandwidth);
}
return y / bandwidth / n;
},
cdf(x) {
let y = 0,
i = 0;
i = 0;
for (; i < n; ++i) {
y += kernel.cdf((x - support[i]) / bandwidth);
}
return y / n;
},
icdf() {
throw Error('KDE icdf not supported.');
}
};

@@ -774,3 +714,2 @@ return dist.data(support);

},
stdev(_) {

@@ -784,3 +723,2 @@ if (arguments.length) {

},
sample: () => sampleLogNormal(mu, sigma),

@@ -796,20 +734,15 @@ pdf: value => densityLogNormal(value, mu, sigma),

let m = 0,
w;
w;
function normalize(x) {
const w = [];
let sum = 0,
i;
i;
for (i = 0; i < m; ++i) {
sum += w[i] = x[i] == null ? 1 : +x[i];
}
for (i = 0; i < m; ++i) {
w[i] /= sum;
}
return w;
}
const dist = {

@@ -821,6 +754,4 @@ weights(_) {

}
return weights;
},
distributions(_) {

@@ -835,15 +766,13 @@ if (arguments.length) {

}
return dist.weights(weights);
}
return dists;
},
sample() {
const r = exports.random();
let d = dists[m - 1],
v = w[0],
i = 0; // first select distribution
v = w[0],
i = 0;
// first select distribution
for (; i < m - 1; v += w[++i]) {

@@ -854,34 +783,25 @@ if (r < v) {

}
} // then sample from it
}
// then sample from it
return d.sample();
},
pdf(x) {
let p = 0,
i = 0;
i = 0;
for (; i < m; ++i) {
p += w[i] * dists[i].pdf(x);
}
return p;
},
cdf(x) {
let p = 0,
i = 0;
i = 0;
for (; i < m; ++i) {
p += w[i] * dists[i].cdf(x);
}
return p;
},
icdf() {
throw Error('Mixture icdf not supported.');
}
};

@@ -896,3 +816,2 @@ return dist.distributions(dists).weights(weights);

}
return min + (max - min) * exports.random();

@@ -905,3 +824,2 @@ }

}
return value >= min && value <= max ? 1 / (max - min) : 0;

@@ -914,3 +832,2 @@ }

}
return value < min ? 0 : value > max ? 1 : (value - min) / (max - min);

@@ -923,3 +840,2 @@ }

}
return p >= 0 && p <= 1 ? min + p * (max - min) : NaN;

@@ -938,3 +854,2 @@ }

},
max(_) {

@@ -948,3 +863,2 @@ if (arguments.length) {

},
sample: () => sampleUniform(a, b),

@@ -955,3 +869,2 @@ pdf: value => densityUniform(value, a, b),

};
if (max == null) {

@@ -961,3 +874,2 @@ max = min == null ? 1 : min;

}
return dist.min(min).max(max);

@@ -969,4 +881,4 @@ }

const delta = uX2 - uX * uX,
slope = Math.abs(delta) < 1e-24 ? 0 : (uXY - uX * uY) / delta,
intercept = uY - slope * uX;
slope = Math.abs(delta) < 1e-24 ? 0 : (uXY - uX * uY) / delta,
intercept = uY - slope * uX;
return [intercept, slope];

@@ -978,21 +890,19 @@ }

let u = x(d),
v = y(d);
v = y(d);
return u != null && (u = +u) >= u && v != null && (v = +v) >= v;
});
if (sort) {
data.sort((a, b) => x(a) - x(b));
}
const n = data.length,
X = new Float64Array(n),
Y = new Float64Array(n); // extract values, calculate means
X = new Float64Array(n),
Y = new Float64Array(n);
// extract values, calculate means
let i = 0,
ux = 0,
uy = 0,
xv,
yv,
d;
ux = 0,
uy = 0,
xv,
yv,
d;
for (d of data) {

@@ -1004,5 +914,5 @@ X[i] = xv = +x(d);

uy += (yv - uy) / i;
} // mean center the data
}
// mean center the data
for (i = 0; i < n; ++i) {

@@ -1012,3 +922,2 @@ X[i] -= ux;

}
return [X, Y, ux, uy];

@@ -1018,9 +927,7 @@ }

let i = -1,
u,
v;
u,
v;
for (const d of data) {
u = x(d);
v = y(d);
if (u != null && (u = +u) >= u && v != null && (v = +v) >= v) {

@@ -1032,10 +939,10 @@ callback(u, v, ++i);

// Adapted from d3-regression by Harry Stevens
// License: https://github.com/HarryStevens/d3-regression/blob/master/LICENSE
function rSquared (data, x, y, uY, predict) {
let SSE = 0,
SST = 0;
SST = 0;
visitPoints(data, x, y, (dx, dy) => {
const sse = dy - predict(dx),
sst = dy - uY;
sst = dy - uY;
SSE += sse * sse;

@@ -1047,10 +954,10 @@ SST += sst * sst;

// Adapted from d3-regression by Harry Stevens
// License: https://github.com/HarryStevens/d3-regression/blob/master/LICENSE
function linear (data, x, y) {
let X = 0,
Y = 0,
XY = 0,
X2 = 0,
n = 0;
Y = 0,
XY = 0,
X2 = 0,
n = 0;
visitPoints(data, x, y, (dx, dy) => {

@@ -1063,6 +970,4 @@ ++n;

});
const coef = ols(X, Y, XY, X2),
predict = x => coef[0] + coef[1] * x;
predict = x => coef[0] + coef[1] * x;
return {

@@ -1075,10 +980,10 @@ coef: coef,

// Adapted from d3-regression by Harry Stevens
// License: https://github.com/HarryStevens/d3-regression/blob/master/LICENSE
function log (data, x, y) {
let X = 0,
Y = 0,
XY = 0,
X2 = 0,
n = 0;
Y = 0,
XY = 0,
X2 = 0,
n = 0;
visitPoints(data, x, y, (dx, dy) => {

@@ -1092,6 +997,4 @@ ++n;

});
const coef = ols(X, Y, XY, X2),
predict = x => coef[0] + coef[1] * Math.log(x);
predict = x => coef[0] + coef[1] * Math.log(x);
return {

@@ -1108,9 +1011,9 @@ coef: coef,

let YL = 0,
XY = 0,
XYL = 0,
X2Y = 0,
n = 0,
dx,
ly,
xy;
XY = 0,
XYL = 0,
X2Y = 0,
n = 0,
dx,
ly,
xy;
visitPoints(data, x, y, (_, dy) => {

@@ -1125,6 +1028,4 @@ dx = xv[n++];

});
const [c0, c1] = ols(XY / uy, YL / uy, XYL / uy, X2Y / uy),
predict = x => Math.exp(c0 + c1 * (x - ux));
predict = x => Math.exp(c0 + c1 * (x - ux));
return {

@@ -1137,14 +1038,14 @@ coef: [Math.exp(c0 - c1 * ux), c1],

// Adapted from d3-regression by Harry Stevens
// License: https://github.com/HarryStevens/d3-regression/blob/master/LICENSE
function pow (data, x, y) {
let X = 0,
Y = 0,
XY = 0,
X2 = 0,
YS = 0,
n = 0;
Y = 0,
XY = 0,
X2 = 0,
YS = 0,
n = 0;
visitPoints(data, x, y, (dx, dy) => {
const lx = Math.log(dx),
ly = Math.log(dy);
ly = Math.log(dy);
++n;

@@ -1157,6 +1058,4 @@ X += (lx - X) / n;

});
const coef = ols(X, Y, XY, X2),
predict = x => coef[0] * Math.pow(x, coef[1]);
predict = x => coef[0] * Math.pow(x, coef[1]);
coef[0] = Math.exp(coef[0]);

@@ -1172,13 +1071,12 @@ return {

const [xv, yv, ux, uy] = points(data, x, y),
n = xv.length;
n = xv.length;
let X2 = 0,
X3 = 0,
X4 = 0,
XY = 0,
X2Y = 0,
i,
dx,
dy,
x2;
X3 = 0,
X4 = 0,
XY = 0,
X2Y = 0,
i,
dx,
dy,
x2;
for (i = 0; i < n;) {

@@ -1194,14 +1092,13 @@ dx = xv[i];

}
const X2X2 = X4 - X2 * X2,
d = X2 * X2X2 - X3 * X3,
a = (X2Y * X2 - XY * X3) / d,
b = (XY * X2X2 - X2Y * X3) / d,
c = -a * X2,
predict = x => {
x = x - ux;
return a * x * x + b * x + c + uy;
}; // transform coefficients back from mean-centered space
d = X2 * X2X2 - X3 * X3,
a = (X2Y * X2 - XY * X3) / d,
b = (XY * X2X2 - X2Y * X3) / d,
c = -a * X2,
predict = x => {
x = x - ux;
return a * x * x + b * x + c + uy;
};
// transform coefficients back from mean-centered space
return {

@@ -1214,2 +1111,3 @@ coef: [c - b * ux + a * ux * ux + uy, b - 2 * a * ux, a],

// Adapted from d3-regression by Harry Stevens
// License: https://github.com/HarryStevens/d3-regression/blob/master/LICENSE

@@ -1219,3 +1117,2 @@ // ... which was adapted from regression-js by Tom Alexander

// License: https://github.com/Tom-Alexander/regression-js/blob/master/LICENSE
function poly (data, x, y, order) {

@@ -1226,8 +1123,7 @@ // use more efficient methods for lower orders

const [xv, yv, ux, uy] = points(data, x, y),
n = xv.length,
lhs = [],
rhs = [],
k = order + 1;
n = xv.length,
lhs = [],
rhs = [],
k = order + 1;
let i, j, l, v, c;
for (i = 0; i < k; ++i) {

@@ -1237,6 +1133,4 @@ for (l = 0, v = 0; l < n; ++l) {

}
lhs.push(v);
c = new Float64Array(k);
for (j = 0; j < k; ++j) {

@@ -1246,21 +1140,14 @@ for (l = 0, v = 0; l < n; ++l) {

}
c[j] = v;
}
rhs.push(c);
}
rhs.push(lhs);
const coef = gaussianElimination(rhs),
predict = x => {
x -= ux;
let y = uy + coef[0] + coef[1] * x + coef[2] * x * x;
for (i = 3; i < k; ++i) y += coef[i] * Math.pow(x, i);
return y;
};
predict = x => {
x -= ux;
let y = uy + coef[0] + coef[1] * x + coef[2] * x * x;
for (i = 3; i < k; ++i) y += coef[i] * Math.pow(x, i);
return y;
};
return {

@@ -1272,10 +1159,10 @@ coef: uncenter(k, coef, -ux, uy),

}
function uncenter(k, a, x, y) {
const z = Array(k);
let i, j, v, c; // initialize to zero
let i, j, v, c;
for (i = 0; i < k; ++i) z[i] = 0; // polynomial expansion
// initialize to zero
for (i = 0; i < k; ++i) z[i] = 0;
// polynomial expansion
for (i = k - 1; i >= 0; --i) {

@@ -1285,25 +1172,21 @@ v = a[i];

z[i] += v;
for (j = 1; j <= i; ++j) {
c *= (i + 1 - j) / j; // binomial coefficent
z[i - j] += v * Math.pow(x, j) * c;
}
} // bias term
}
// bias term
z[0] += y;
return z;
} // Given an array for a two-dimensional matrix and the polynomial order,
}
// Given an array for a two-dimensional matrix and the polynomial order,
// solve A * x = b using Gaussian elimination.
function gaussianElimination(matrix) {
const n = matrix.length - 1,
coef = [];
coef = [];
let i, j, k, r, t;
for (i = 0; i < n; ++i) {
r = i; // max row
for (j = i + 1; j < n; ++j) {

@@ -1314,3 +1197,2 @@ if (Math.abs(matrix[i][j]) > Math.abs(matrix[i][r])) {

}
for (k = i; k < n + 1; ++k) {

@@ -1321,3 +1203,2 @@ t = matrix[k][i];

}
for (j = i + 1; j < n; ++j) {

@@ -1329,13 +1210,9 @@ for (k = n; k >= i; k--) {

}
for (j = n - 1; j >= 0; --j) {
t = 0;
for (k = j + 1; k < n; ++k) {
t += matrix[k][j] * coef[k];
}
coef[j] = (matrix[n][j] - t) / matrix[j][j];
}
return coef;

@@ -1345,28 +1222,27 @@ }

const maxiters = 2,
epsilon = 1e-12; // Adapted from science.js by Jason Davies
epsilon = 1e-12;
// Adapted from science.js by Jason Davies
// Source: https://github.com/jasondavies/science.js/blob/master/src/stats/loess.js
// License: https://github.com/jasondavies/science.js/blob/master/LICENSE
function loess (data, x, y, bandwidth) {
const [xv, yv, ux, uy] = points(data, x, y, true),
n = xv.length,
bw = Math.max(2, ~~(bandwidth * n)),
// # nearest neighbors
yhat = new Float64Array(n),
residuals = new Float64Array(n),
robustWeights = new Float64Array(n).fill(1);
n = xv.length,
bw = Math.max(2, ~~(bandwidth * n)),
// # nearest neighbors
yhat = new Float64Array(n),
residuals = new Float64Array(n),
robustWeights = new Float64Array(n).fill(1);
for (let iter = -1; ++iter <= maxiters;) {
const interval = [0, bw - 1];
for (let i = 0; i < n; ++i) {
const dx = xv[i],
i0 = interval[0],
i1 = interval[1],
edge = dx - xv[i0] > xv[i1] - dx ? i0 : i1;
i0 = interval[0],
i1 = interval[1],
edge = dx - xv[i0] > xv[i1] - dx ? i0 : i1;
let W = 0,
X = 0,
Y = 0,
XY = 0,
X2 = 0;
X = 0,
Y = 0,
XY = 0,
X2 = 0;
const denom = 1 / Math.abs(xv[edge] - dx || 1); // avoid singularity!

@@ -1376,5 +1252,5 @@

const xk = xv[k],
yk = yv[k],
w = tricube(Math.abs(dx - xk) * denom) * robustWeights[k],
xkw = xk * w;
yk = yv[k],
w = tricube(Math.abs(dx - xk) * denom) * robustWeights[k],
xkw = xk * w;
W += w;

@@ -1385,5 +1261,5 @@ X += xkw;

X2 += xk * xkw;
} // linear regression fit
}
// linear regression fit
const [a, b] = ols(X / W, Y / W, XY / W, X2 / W);

@@ -1394,33 +1270,31 @@ yhat[i] = a + b * dx;

}
if (iter === maxiters) {
break;
}
const medianResidual = median(residuals);
if (Math.abs(medianResidual) < epsilon) break;
for (let i = 0, arg, w; i < n; ++i) {
arg = residuals[i] / (6 * medianResidual); // default to epsilon (rather than zero) for large deviations
arg = residuals[i] / (6 * medianResidual);
// default to epsilon (rather than zero) for large deviations
// keeping weights tiny but non-zero prevents singularites
robustWeights[i] = arg >= 1 ? epsilon : (w = 1 - arg * arg) * w;
}
}
return output(xv, yhat, ux, uy);
} // weighting kernel for local regression
}
// weighting kernel for local regression
function tricube(x) {
return (x = 1 - x * x * x) * x * x;
} // advance sliding window interval of nearest neighbors
}
// advance sliding window interval of nearest neighbors
function updateInterval(xv, i, interval) {
const val = xv[i];
let left = interval[0],
right = interval[1] + 1;
if (right >= xv.length) return; // step right if distance to new right edge is <= distance to old left edge
right = interval[1] + 1;
if (right >= xv.length) return;
// step right if distance to new right edge is <= distance to old left edge
// step when distance is equal to ensure movement over duplicate x values
while (i > left && xv[right] - val <= val - xv[left]) {

@@ -1431,17 +1305,15 @@ interval[0] = ++left;

}
} // generate smoothed output points
}
// generate smoothed output points
// average points with repeated x values
function output(xv, yhat, ux, uy) {
const n = xv.length,
out = [];
out = [];
let i = 0,
cnt = 0,
prev = [],
v;
cnt = 0,
prev = [],
v;
for (; i < n; ++i) {
v = xv[i] + ux;
if (prev[0] === v) {

@@ -1458,3 +1330,2 @@ // average output values via online update

}
prev[1] += uy;

@@ -1465,16 +1336,15 @@ return out;

// subdivide up to accuracy of 0.5 degrees
const MIN_RADIANS = 0.5 * Math.PI / 180; // Adaptively sample an interpolated function over a domain extent
const MIN_RADIANS = 0.5 * Math.PI / 180;
// Adaptively sample an interpolated function over a domain extent
function sampleCurve (f, extent, minSteps, maxSteps) {
minSteps = minSteps || 25;
maxSteps = Math.max(minSteps, maxSteps || 200);
const point = x => [x, f(x)],
minX = extent[0],
maxX = extent[1],
span = maxX - minX,
stop = span / maxSteps,
prev = [point(minX)],
next = [];
minX = extent[0],
maxX = extent[1],
span = maxX - minX,
stop = span / maxSteps,
prev = [point(minX)],
next = [];
if (minSteps === maxSteps) {

@@ -1485,3 +1355,2 @@ // no adaptation, sample uniform grid directly and return

}
prev.push(point(maxX));

@@ -1493,3 +1362,2 @@ return prev;

next.push(point(maxX));
for (let i = minSteps; --i > 0;) {

@@ -1499,3 +1367,2 @@ next.push(point(minX + i / minSteps * span));

}
let p0 = prev[0];

@@ -1505,3 +1372,2 @@ let p1 = next[next.length - 1];

const sy = scaleY(p0[1], next);
while (p1) {

@@ -1511,3 +1377,2 @@ // midpoint for potential curve subdivision

const dx = pm[0] - p0[0] >= stop;
if (dx && angleDelta(p0, pm, p1, sx, sy) > MIN_RADIANS) {

@@ -1525,9 +1390,6 @@ // maximum resolution has not yet been met, and

}
p1 = next[next.length - 1];
}
return prev;
}
function scaleY(init, points) {

@@ -1537,3 +1399,2 @@ let ymin = init;

const n = points.length;
for (let i = 0; i < n; ++i) {

@@ -1544,9 +1405,7 @@ const y = points[i][1];

}
return 1 / (ymax - ymin);
}
function angleDelta(p, q, r, sx, sy) {
const a0 = Math.atan2(sy * (r[1] - p[1]), sx * (r[0] - p[0])),
a1 = Math.atan2(sy * (q[1] - p[1]), sx * (q[0] - p[0]));
a1 = Math.atan2(sy * (q[1] - p[1]), sx * (q[0] - p[0]));
return Math.abs(a0 - a1);

@@ -1590,4 +1449,2 @@ }

Object.defineProperty(exports, '__esModule', { value: true });
}));

2

build/vega-statistics.min.js

@@ -1,2 +0,2 @@

!function(t,n){"object"==typeof exports&&"undefined"!=typeof module?n(exports):"function"==typeof define&&define.amd?define(["exports"],n):n((t="undefined"!=typeof globalThis?globalThis:t||self).vega={})}(this,(function(t){"use strict";function*n(t,n){if(null==n)for(let n of t)null!=n&&""!==n&&(n=+n)>=n&&(yield n);else{let e=-1;for(let r of t)r=n(r,++e,t),null!=r&&""!==r&&(r=+r)>=r&&(yield r)}}function e(t,n){return null==t||null==n?NaN:t<n?-1:t>n?1:t>=n?0:NaN}function r(t){return null===t?NaN:+t}function o(){let t=arguments.length>0&&void 0!==arguments[0]?arguments[0]:e;if(t===e)return l;if("function"!=typeof t)throw new TypeError("compare is not a function");return(n,e)=>{const r=t(n,e);return r||0===r?r:(0===t(e,e))-(0===t(n,n))}}function l(t,n){return(null==t||!(t>=t))-(null==n||!(n>=n))||(t<n?-1:t>n?1:0)}function u(t,n){let e;if(void 0===n)for(const n of t)null!=n&&(e<n||void 0===e&&n>=n)&&(e=n);else{let r=-1;for(let o of t)null!=(o=n(o,++r,t))&&(e<o||void 0===e&&o>=o)&&(e=o)}return e}function f(t,n){let e;if(void 0===n)for(const n of t)null!=n&&(e>n||void 0===e&&n>=n)&&(e=n);else{let r=-1;for(let o of t)null!=(o=n(o,++r,t))&&(e>o||void 0===e&&o>=o)&&(e=o)}return e}function a(t,n){let e=arguments.length>2&&void 0!==arguments[2]?arguments[2]:0,r=arguments.length>3&&void 0!==arguments[3]?arguments[3]:t.length-1,u=arguments.length>4?arguments[4]:void 0;for(u=void 0===u?l:o(u);r>e;){if(r-e>600){const o=r-e+1,l=n-e+1,f=Math.log(o),i=.5*Math.exp(2*f/3),s=.5*Math.sqrt(f*i*(o-i)/o)*(l-o/2<0?-1:1);a(t,n,Math.max(e,Math.floor(n-l*i/o+s)),Math.min(r,Math.floor(n+(o-l)*i/o+s)),u)}const o=t[n];let l=e,f=r;for(i(t,e,n),u(t[r],o)>0&&i(t,e,r);l<f;){for(i(t,l,f),++l,--f;u(t[l],o)<0;)++l;for(;u(t[f],o)>0;)--f}0===u(t[e],o)?i(t,e,f):(++f,i(t,f,r)),f<=n&&(e=f+1),n<=f&&(r=f-1)}return t}function i(t,n,e){const r=t[n];t[n]=t[e],t[e]=r}function s(t,n,e){if(t=Float64Array.from(function*(t,n){if(void 0===n)for(let n of t)null!=n&&(n=+n)>=n&&(yield n);else{let e=-1;for(let r of t)null!=(r=n(r,++e,t))&&(r=+r)>=r&&(yield r)}}(t,e)),r=t.length){if((n=+n)<=0||r<2)return f(t);if(n>=1)return u(t);var r,o=(r-1)*n,l=Math.floor(o),i=u(a(t,l).subarray(0,l+1));return i+(f(t.subarray(l+1))-i)*(o-l)}}function c(t,o,l){const u=Float64Array.from(n(t,l));return u.sort(e),o.map((t=>function(t,n){let e=arguments.length>2&&void 0!==arguments[2]?arguments[2]:r;if(o=t.length){if((n=+n)<=0||o<2)return+e(t[0],0,t);if(n>=1)return+e(t[o-1],o-1,t);var o,l=(o-1)*n,u=Math.floor(l),f=+e(t[u],u,t);return f+(+e(t[u+1],u+1,t)-f)*(l-u)}}(u,t)))}function h(t,n){return c(t,[.25,.5,.75],n)}function d(t,n){const e=t.length,r=function(t,n){const e=function(t,n){let e,r=0,o=0,l=0;if(void 0===n)for(let n of t)null!=n&&(n=+n)>=n&&(e=n-o,o+=e/++r,l+=e*(n-o));else{let u=-1;for(let f of t)null!=(f=n(f,++u,t))&&(f=+f)>=f&&(e=f-o,o+=e/++r,l+=e*(f-o))}if(r>1)return l/(r-1)}(t,n);return e?Math.sqrt(e):e}(t,n),o=h(t,n),l=(o[2]-o[0])/1.34;return 1.06*(Math.min(r,l)||r||Math.abs(o[0])||1)*Math.pow(e,-.2)}t.random=Math.random;const p=Math.sqrt(2*Math.PI),m=Math.SQRT2;let M=NaN;function g(n,e){n=n||0,e=null==e?1:e;let r,o,l=0,u=0;if(M==M)l=M,M=NaN;else{do{l=2*t.random()-1,u=2*t.random()-1,r=l*l+u*u}while(0===r||r>1);o=Math.sqrt(-2*Math.log(r)/r),l*=o,M=u*o}return n+l*e}function b(t,n,e){const r=(t-(n||0))/(e=null==e?1:e);return Math.exp(-.5*r*r)/(e*p)}function v(t,n,e){const r=(t-(n=n||0))/(e=null==e?1:e),o=Math.abs(r);let l;if(o>37)l=0;else{const t=Math.exp(-o*o/2);let n;o<7.07106781186547?(n=.0352624965998911*o+.700383064443688,n=n*o+6.37396220353165,n=n*o+33.912866078383,n=n*o+112.079291497871,n=n*o+221.213596169931,n=n*o+220.206867912376,l=t*n,n=.0883883476483184*o+1.75566716318264,n=n*o+16.064177579207,n=n*o+86.7807322029461,n=n*o+296.564248779674,n=n*o+637.333633378831,n=n*o+793.826512519948,n=n*o+440.413735824752,l/=n):(n=o+.65,n=o+4/n,n=o+3/n,n=o+2/n,n=o+1/n,l=t/n/2.506628274631)}return r>0?1-l:l}function y(t,n,e){return t<0||t>1?NaN:(n||0)+(null==e?1:e)*m*function(t){let n,e=-Math.log((1-t)*(1+t));e<6.25?(e-=3.125,n=-364441206401782e-35,n=n*e-16850591381820166e-35,n=128584807152564e-32+n*e,n=11157877678025181e-33+n*e,n=n*e-1333171662854621e-31,n=20972767875968562e-33+n*e,n=6637638134358324e-30+n*e,n=n*e-4054566272975207e-29,n=n*e-8151934197605472e-29,n=26335093153082323e-28+n*e,n=n*e-12975133253453532e-27,n=n*e-5415412054294628e-26,n=1.0512122733215323e-9+n*e,n=n*e-4.112633980346984e-9,n=n*e-2.9070369957882005e-8,n=4.2347877827932404e-7+n*e,n=n*e-13654692000834679e-22,n=n*e-13882523362786469e-21,n=.00018673420803405714+n*e,n=n*e-.000740702534166267,n=n*e-.006033670871430149,n=.24015818242558962+n*e,n=1.6536545626831027+n*e):e<16?(e=Math.sqrt(e)-3.25,n=2.2137376921775787e-9,n=9.075656193888539e-8+n*e,n=n*e-2.7517406297064545e-7,n=1.8239629214389228e-8+n*e,n=15027403968909828e-22+n*e,n=n*e-4013867526981546e-21,n=29234449089955446e-22+n*e,n=12475304481671779e-21+n*e,n=n*e-47318229009055734e-21,n=6828485145957318e-20+n*e,n=24031110387097894e-21+n*e,n=n*e-.0003550375203628475,n=.0009532893797373805+n*e,n=n*e-.0016882755560235047,n=.002491442096107851+n*e,n=n*e-.003751208507569241,n=.005370914553590064+n*e,n=1.0052589676941592+n*e,n=3.0838856104922208+n*e):Number.isFinite(e)?(e=Math.sqrt(e)-5,n=-27109920616438573e-27,n=n*e-2.555641816996525e-10,n=1.5076572693500548e-9+n*e,n=n*e-3.789465440126737e-9,n=7.61570120807834e-9+n*e,n=n*e-1.496002662714924e-8,n=2.914795345090108e-8+n*e,n=n*e-6.771199775845234e-8,n=2.2900482228026655e-7+n*e,n=n*e-9.9298272942317e-7,n=4526062597223154e-21+n*e,n=n*e-1968177810553167e-20,n=7599527703001776e-20+n*e,n=n*e-.00021503011930044477,n=n*e-.00013871931833623122,n=1.0103004648645344+n*e,n=4.849906401408584+n*e):n=1/0;return n*t}(2*t-1)}function N(t,n){let e,r;const o={mean(t){return arguments.length?(e=t||0,o):e},stdev(t){return arguments.length?(r=null==t?1:t,o):r},sample:()=>g(e,r),pdf:t=>b(t,e,r),cdf:t=>v(t,e,r),icdf:t=>y(t,e,r)};return o.mean(t).stdev(n)}function w(t,n){return t=t||0,n=null==n?1:n,Math.exp(t+g()*n)}function x(t,n,e){if(t<=0)return 0;n=n||0,e=null==e?1:e;const r=(Math.log(t)-n)/e;return Math.exp(-.5*r*r)/(e*p*t)}function q(t,n,e){return v(Math.log(t),n,e)}function A(t,n,e){return Math.exp(y(t,n,e))}function F(n,e){return null==e&&(e=null==n?1:n,n=0),n+(e-n)*t.random()}function L(t,n,e){return null==e&&(e=null==n?1:n,n=0),t>=n&&t<=e?1/(e-n):0}function S(t,n,e){return null==e&&(e=null==n?1:n,n=0),t<n?0:t>e?1:(t-n)/(e-n)}function E(t,n,e){return null==e&&(e=null==n?1:n,n=0),t>=0&&t<=1?n+t*(e-n):NaN}function P(t,n,e,r){const o=r-t*t,l=Math.abs(o)<1e-24?0:(e-t*n)/o;return[n-l*t,l]}function U(t,n,e,r){t=t.filter((t=>{let r=n(t),o=e(t);return null!=r&&(r=+r)>=r&&null!=o&&(o=+o)>=o})),r&&t.sort(((t,e)=>n(t)-n(e)));const o=t.length,l=new Float64Array(o),u=new Float64Array(o);let f,a,i,s=0,c=0,h=0;for(i of t)l[s]=f=+n(i),u[s]=a=+e(i),++s,c+=(f-c)/s,h+=(a-h)/s;for(s=0;s<o;++s)l[s]-=c,u[s]-=h;return[l,u,c,h]}function I(t,n,e,r){let o,l,u=-1;for(const f of t)o=n(f),l=e(f),null!=o&&(o=+o)>=o&&null!=l&&(l=+l)>=l&&r(o,l,++u)}function T(t,n,e,r,o){let l=0,u=0;return I(t,n,e,((t,n)=>{const e=n-o(t),f=n-r;l+=e*e,u+=f*f})),1-l/u}function k(t,n,e){let r=0,o=0,l=0,u=0,f=0;I(t,n,e,((t,n)=>{++f,r+=(t-r)/f,o+=(n-o)/f,l+=(t*n-l)/f,u+=(t*t-u)/f}));const a=P(r,o,l,u),i=t=>a[0]+a[1]*t;return{coef:a,predict:i,rSquared:T(t,n,e,o,i)}}function C(t,n,e){const[r,o,l,u]=U(t,n,e),f=r.length;let a,i,s,c,h=0,d=0,p=0,m=0,M=0;for(a=0;a<f;)i=r[a],s=o[a++],c=i*i,h+=(c-h)/a,d+=(c*i-d)/a,p+=(c*c-p)/a,m+=(i*s-m)/a,M+=(c*s-M)/a;const g=p-h*h,b=h*g-d*d,v=(M*h-m*d)/b,y=(m*g-M*d)/b,N=-v*h,w=t=>v*(t-=l)*t+y*t+N+u;return{coef:[N-y*l+v*l*l+u,y-2*v*l,v],predict:w,rSquared:T(t,n,e,u,w)}}function D(t,n,e,r){const o=Array(t);let l,u,f,a;for(l=0;l<t;++l)o[l]=0;for(l=t-1;l>=0;--l)for(f=n[l],a=1,o[l]+=f,u=1;u<=l;++u)a*=(l+1-u)/u,o[l-u]+=f*Math.pow(e,u)*a;return o[0]+=r,o}function R(t){return(t=1-t*t*t)*t*t}function j(t,n,e){const r=t[n];let o=e[0],l=e[1]+1;if(!(l>=t.length))for(;n>o&&t[l]-r<=r-t[o];)e[0]=++o,e[1]=l,++l}const K=.5*Math.PI/180;function Q(t,n,e,r,o){const l=Math.atan2(o*(e[1]-t[1]),r*(e[0]-t[0])),u=Math.atan2(o*(n[1]-t[1]),r*(n[0]-t[0]));return Math.abs(l-u)}t.bandwidthNRD=d,t.bin=function(t){const n=t.maxbins||20,e=t.base||10,r=Math.log(e),o=t.divide||[5,2];let l,u,f,a,i,s,c=t.extent[0],h=t.extent[1];const d=t.span||h-c||Math.abs(c)||1;if(t.step)l=t.step;else if(t.steps){for(a=d/n,i=0,s=t.steps.length;i<s&&t.steps[i]<a;++i);l=t.steps[Math.max(0,i-1)]}else{for(u=Math.ceil(Math.log(n)/r),f=t.minstep||0,l=Math.max(f,Math.pow(e,Math.round(Math.log(d)/r)-u));Math.ceil(d/l)>n;)l*=e;for(i=0,s=o.length;i<s;++i)a=l/o[i],a>=f&&d/a<=n&&(l=a)}a=Math.log(l);const p=a>=0?0:1+~~(-a/r),m=Math.pow(e,-p-1);return(t.nice||void 0===t.nice)&&(a=Math.floor(c/l+m)*l,c=c<a?a-l:a,h=Math.ceil(h/l)*l),{start:c,stop:h===c?c+l:h,step:l}},t.bootstrapCI=function(r,o,l,u){if(!r.length)return[void 0,void 0];const f=Float64Array.from(n(r,u)),a=f.length,i=o;let c,h,d,p;for(d=0,p=Array(i);d<i;++d){for(c=0,h=0;h<a;++h)c+=f[~~(t.random()*a)];p[d]=c/a}return p.sort(e),[s(p,l/2),s(p,1-l/2)]},t.cumulativeLogNormal=q,t.cumulativeNormal=v,t.cumulativeUniform=S,t.densityLogNormal=x,t.densityNormal=b,t.densityUniform=L,t.dotbin=function(t,n,e,r){r=r||(t=>t);const o=t.length,l=new Float64Array(o);let u,f=0,a=1,i=r(t[0]),s=i,c=i+n;for(;a<o;++a){if(u=r(t[a]),u>=c){for(s=(i+s)/2;f<a;++f)l[f]=s;c=u+n,i=u}s=u}for(s=(i+s)/2;f<a;++f)l[f]=s;return e?function(t,n){const e=t.length;let r,o,l=0,u=1;for(;t[l]===t[u];)++u;for(;u<e;){for(r=u+1;t[u]===t[r];)++r;if(t[u]-t[u-1]<n){for(o=u+(l+r-u-u>>1);o<u;)t[o++]=t[u];for(;o>u;)t[o--]=t[l]}l=u,u=r}return t}(l,n+n/4):l},t.quantileLogNormal=A,t.quantileNormal=y,t.quantileUniform=E,t.quantiles=c,t.quartiles=h,t.randomInteger=function(n,e){let r,o,l;null==e&&(e=n,n=0);const u={min(t){return arguments.length?(r=t||0,l=o-r,u):r},max(t){return arguments.length?(o=t||0,l=o-r,u):o},sample:()=>r+Math.floor(l*t.random()),pdf:t=>t===Math.floor(t)&&t>=r&&t<o?1/l:0,cdf(t){const n=Math.floor(t);return n<r?0:n>=o?1:(n-r+1)/l},icdf:t=>t>=0&&t<=1?r-1+Math.floor(t*l):NaN};return u.min(n).max(e)},t.randomKDE=function(n,e){const r=N();let o=0;const l={data(t){return arguments.length?(n=t,o=t?t.length:0,l.bandwidth(e)):n},bandwidth(t){return arguments.length?(!(e=t)&&n&&(e=d(n)),l):e},sample:()=>n[~~(t.random()*o)]+e*r.sample(),pdf(t){let l=0,u=0;for(;u<o;++u)l+=r.pdf((t-n[u])/e);return l/e/o},cdf(t){let l=0,u=0;for(;u<o;++u)l+=r.cdf((t-n[u])/e);return l/o},icdf(){throw Error("KDE icdf not supported.")}};return l.data(n)},t.randomLCG=function(t){return function(){return(t=(1103515245*t+12345)%2147483647)/2147483647}},t.randomLogNormal=function(t,n){let e,r;const o={mean(t){return arguments.length?(e=t||0,o):e},stdev(t){return arguments.length?(r=null==t?1:t,o):r},sample:()=>w(e,r),pdf:t=>x(t,e,r),cdf:t=>q(t,e,r),icdf:t=>A(t,e,r)};return o.mean(t).stdev(n)},t.randomMixture=function(n,e){let r,o=0;const l={weights(t){return arguments.length?(r=function(t){const n=[];let e,r=0;for(e=0;e<o;++e)r+=n[e]=null==t[e]?1:+t[e];for(e=0;e<o;++e)n[e]/=r;return n}(e=t||[]),l):e},distributions(t){return arguments.length?(t?(o=t.length,n=t):(o=0,n=[]),l.weights(e)):n},sample(){const e=t.random();let l=n[o-1],u=r[0],f=0;for(;f<o-1;u+=r[++f])if(e<u){l=n[f];break}return l.sample()},pdf(t){let e=0,l=0;for(;l<o;++l)e+=r[l]*n[l].pdf(t);return e},cdf(t){let e=0,l=0;for(;l<o;++l)e+=r[l]*n[l].cdf(t);return e},icdf(){throw Error("Mixture icdf not supported.")}};return l.distributions(n).weights(e)},t.randomNormal=N,t.randomUniform=function(t,n){let e,r;const o={min(t){return arguments.length?(e=t||0,o):e},max(t){return arguments.length?(r=null==t?1:t,o):r},sample:()=>F(e,r),pdf:t=>L(t,e,r),cdf:t=>S(t,e,r),icdf:t=>E(t,e,r)};return null==n&&(n=null==t?1:t,t=0),o.min(t).max(n)},t.regressionExp=function(t,n,e){const[r,o,l,u]=U(t,n,e);let f,a,i,s=0,c=0,h=0,d=0,p=0;I(t,n,e,((t,n)=>{f=r[p++],a=Math.log(n),i=f*n,s+=(n*a-s)/p,c+=(i-c)/p,h+=(i*a-h)/p,d+=(f*i-d)/p}));const[m,M]=P(c/u,s/u,h/u,d/u),g=t=>Math.exp(m+M*(t-l));return{coef:[Math.exp(m-M*l),M],predict:g,rSquared:T(t,n,e,u,g)}},t.regressionLinear=k,t.regressionLoess=function(t,n,e,r){const[o,l,u,f]=U(t,n,e,!0),a=o.length,i=Math.max(2,~~(r*a)),c=new Float64Array(a),h=new Float64Array(a),d=new Float64Array(a).fill(1);for(let t=-1;++t<=2;){const n=[0,i-1];for(let t=0;t<a;++t){const e=o[t],r=n[0],u=n[1],f=e-o[r]>o[u]-e?r:u;let a=0,i=0,s=0,p=0,m=0;const M=1/Math.abs(o[f]-e||1);for(let t=r;t<=u;++t){const n=o[t],r=l[t],u=R(Math.abs(e-n)*M)*d[t],f=n*u;a+=u,i+=f,s+=r*u,p+=r*f,m+=n*f}const[g,b]=P(i/a,s/a,p/a,m/a);c[t]=g+b*e,h[t]=Math.abs(l[t]-c[t]),j(o,t+1,n)}if(2===t)break;const e=s(h,.5,void 0);if(Math.abs(e)<1e-12)break;for(let t,n,r=0;r<a;++r)t=h[r]/(6*e),d[r]=t>=1?1e-12:(n=1-t*t)*n}return function(t,n,e,r){const o=t.length,l=[];let u,f=0,a=0,i=[];for(;f<o;++f)u=t[f]+e,i[0]===u?i[1]+=(n[f]-i[1])/++a:(a=0,i[1]+=r,i=[u,n[f]],l.push(i));return i[1]+=r,l}(o,c,u,f)},t.regressionLog=function(t,n,e){let r=0,o=0,l=0,u=0,f=0;I(t,n,e,((t,n)=>{++f,t=Math.log(t),r+=(t-r)/f,o+=(n-o)/f,l+=(t*n-l)/f,u+=(t*t-u)/f}));const a=P(r,o,l,u),i=t=>a[0]+a[1]*Math.log(t);return{coef:a,predict:i,rSquared:T(t,n,e,o,i)}},t.regressionPoly=function(t,n,e,r){if(1===r)return k(t,n,e);if(2===r)return C(t,n,e);const[o,l,u,f]=U(t,n,e),a=o.length,i=[],s=[],c=r+1;let h,d,p,m,M;for(h=0;h<c;++h){for(p=0,m=0;p<a;++p)m+=Math.pow(o[p],h)*l[p];for(i.push(m),M=new Float64Array(c),d=0;d<c;++d){for(p=0,m=0;p<a;++p)m+=Math.pow(o[p],h+d);M[d]=m}s.push(M)}s.push(i);const g=function(t){const n=t.length-1,e=[];let r,o,l,u,f;for(r=0;r<n;++r){for(u=r,o=r+1;o<n;++o)Math.abs(t[r][o])>Math.abs(t[r][u])&&(u=o);for(l=r;l<n+1;++l)f=t[l][r],t[l][r]=t[l][u],t[l][u]=f;for(o=r+1;o<n;++o)for(l=n;l>=r;l--)t[l][o]-=t[l][r]*t[r][o]/t[r][r]}for(o=n-1;o>=0;--o){for(f=0,l=o+1;l<n;++l)f+=t[l][o]*e[l];e[o]=(t[n][o]-f)/t[o][o]}return e}(s),b=t=>{t-=u;let n=f+g[0]+g[1]*t+g[2]*t*t;for(h=3;h<c;++h)n+=g[h]*Math.pow(t,h);return n};return{coef:D(c,g,-u,f),predict:b,rSquared:T(t,n,e,f,b)}},t.regressionPow=function(t,n,e){let r=0,o=0,l=0,u=0,f=0,a=0;I(t,n,e,((t,n)=>{const e=Math.log(t),i=Math.log(n);++a,r+=(e-r)/a,o+=(i-o)/a,l+=(e*i-l)/a,u+=(e*e-u)/a,f+=(n-f)/a}));const i=P(r,o,l,u),s=t=>i[0]*Math.pow(t,i[1]);return i[0]=Math.exp(i[0]),{coef:i,predict:s,rSquared:T(t,n,e,f,s)}},t.regressionQuad=C,t.sampleCurve=function(t,n,e,r){e=e||25,r=Math.max(e,r||200);const o=n=>[n,t(n)],l=n[0],u=n[1],f=u-l,a=f/r,i=[o(l)],s=[];if(e===r){for(let t=1;t<r;++t)i.push(o(l+t/e*f));return i.push(o(u)),i}s.push(o(u));for(let t=e;--t>0;)s.push(o(l+t/e*f));let c=i[0],h=s[s.length-1];const d=1/f,p=function(t,n){let e=t,r=t;const o=n.length;for(let t=0;t<o;++t){const o=n[t][1];o<e&&(e=o),o>r&&(r=o)}return 1/(r-e)}(c[1],s);for(;h;){const t=o((c[0]+h[0])/2);t[0]-c[0]>=a&&Q(c,t,h,d,p)>K?s.push(t):(c=h,i.push(h),s.pop()),h=s[s.length-1]}return i},t.sampleLogNormal=w,t.sampleNormal=g,t.sampleUniform=F,t.setRandom=function(n){t.random=n},Object.defineProperty(t,"__esModule",{value:!0})}));
!function(t,n){"object"==typeof exports&&"undefined"!=typeof module?n(exports):"function"==typeof define&&define.amd?define(["exports"],n):n((t="undefined"!=typeof globalThis?globalThis:t||self).vega={})}(this,(function(t){"use strict";function*n(t,n){if(null==n)for(let n of t)null!=n&&""!==n&&(n=+n)>=n&&(yield n);else{let e=-1;for(let r of t)r=n(r,++e,t),null!=r&&""!==r&&(r=+r)>=r&&(yield r)}}function e(t,n){return null==t||null==n?NaN:t<n?-1:t>n?1:t>=n?0:NaN}function r(t){return null===t?NaN:+t}function o(t,n){return(null==t||!(t>=t))-(null==n||!(n>=n))||(t<n?-1:t>n?1:0)}function l(t,n){let e;if(void 0===n)for(const n of t)null!=n&&(e<n||void 0===e&&n>=n)&&(e=n);else{let r=-1;for(let o of t)null!=(o=n(o,++r,t))&&(e<o||void 0===e&&o>=o)&&(e=o)}return e}function u(t,n){let e;if(void 0===n)for(const n of t)null!=n&&(e>n||void 0===e&&n>=n)&&(e=n);else{let r=-1;for(let o of t)null!=(o=n(o,++r,t))&&(e>o||void 0===e&&o>=o)&&(e=o)}return e}function f(t,n){let r=arguments.length>2&&void 0!==arguments[2]?arguments[2]:0,l=arguments.length>3&&void 0!==arguments[3]?arguments[3]:1/0,u=arguments.length>4?arguments[4]:void 0;if(n=Math.floor(n),r=Math.floor(Math.max(0,r)),l=Math.floor(Math.min(t.length-1,l)),!(r<=n&&n<=l))return t;for(u=void 0===u?o:function(){let t=arguments.length>0&&void 0!==arguments[0]?arguments[0]:e;if(t===e)return o;if("function"!=typeof t)throw new TypeError("compare is not a function");return(n,e)=>{const r=t(n,e);return r||0===r?r:(0===t(e,e))-(0===t(n,n))}}(u);l>r;){if(l-r>600){const e=l-r+1,o=n-r+1,a=Math.log(e),i=.5*Math.exp(2*a/3),s=.5*Math.sqrt(a*i*(e-i)/e)*(o-e/2<0?-1:1);f(t,n,Math.max(r,Math.floor(n-o*i/e+s)),Math.min(l,Math.floor(n+(e-o)*i/e+s)),u)}const e=t[n];let o=r,i=l;for(a(t,r,n),u(t[l],e)>0&&a(t,r,l);o<i;){for(a(t,o,i),++o,--i;u(t[o],e)<0;)++o;for(;u(t[i],e)>0;)--i}0===u(t[r],e)?a(t,r,i):(++i,a(t,i,l)),i<=n&&(r=i+1),n<=i&&(l=i-1)}return t}function a(t,n,e){const r=t[n];t[n]=t[e],t[e]=r}function i(t,n,e){if(t=Float64Array.from(function*(t,n){if(void 0===n)for(let n of t)null!=n&&(n=+n)>=n&&(yield n);else{let e=-1;for(let r of t)null!=(r=n(r,++e,t))&&(r=+r)>=r&&(yield r)}}(t,e)),(r=t.length)&&!isNaN(n=+n)){if(n<=0||r<2)return u(t);if(n>=1)return l(t);var r,o=(r-1)*n,a=Math.floor(o),i=l(f(t,a).subarray(0,a+1));return i+(u(t.subarray(a+1))-i)*(o-a)}}function s(t,o,l){const u=Float64Array.from(n(t,l));return u.sort(e),o.map((t=>function(t,n){let e=arguments.length>2&&void 0!==arguments[2]?arguments[2]:r;if((o=t.length)&&!isNaN(n=+n)){if(n<=0||o<2)return+e(t[0],0,t);if(n>=1)return+e(t[o-1],o-1,t);var o,l=(o-1)*n,u=Math.floor(l),f=+e(t[u],u,t);return f+(+e(t[u+1],u+1,t)-f)*(l-u)}}(u,t)))}function c(t,n){return s(t,[.25,.5,.75],n)}function h(t,n){const e=t.length,r=function(t,n){const e=function(t,n){let e,r=0,o=0,l=0;if(void 0===n)for(let n of t)null!=n&&(n=+n)>=n&&(e=n-o,o+=e/++r,l+=e*(n-o));else{let u=-1;for(let f of t)null!=(f=n(f,++u,t))&&(f=+f)>=f&&(e=f-o,o+=e/++r,l+=e*(f-o))}if(r>1)return l/(r-1)}(t,n);return e?Math.sqrt(e):e}(t,n),o=c(t,n),l=(o[2]-o[0])/1.34;return 1.06*(Math.min(r,l)||r||Math.abs(o[0])||1)*Math.pow(e,-.2)}t.random=Math.random;const d=Math.sqrt(2*Math.PI),p=Math.SQRT2;let M=NaN;function m(n,e){n=n||0,e=null==e?1:e;let r,o,l=0,u=0;if(M==M)l=M,M=NaN;else{do{l=2*t.random()-1,u=2*t.random()-1,r=l*l+u*u}while(0===r||r>1);o=Math.sqrt(-2*Math.log(r)/r),l*=o,M=u*o}return n+l*e}function g(t,n,e){const r=(t-(n||0))/(e=null==e?1:e);return Math.exp(-.5*r*r)/(e*d)}function N(t,n,e){const r=(t-(n=n||0))/(e=null==e?1:e),o=Math.abs(r);let l;if(o>37)l=0;else{const t=Math.exp(-o*o/2);let n;o<7.07106781186547?(n=.0352624965998911*o+.700383064443688,n=n*o+6.37396220353165,n=n*o+33.912866078383,n=n*o+112.079291497871,n=n*o+221.213596169931,n=n*o+220.206867912376,l=t*n,n=.0883883476483184*o+1.75566716318264,n=n*o+16.064177579207,n=n*o+86.7807322029461,n=n*o+296.564248779674,n=n*o+637.333633378831,n=n*o+793.826512519948,n=n*o+440.413735824752,l/=n):(n=o+.65,n=o+4/n,n=o+3/n,n=o+2/n,n=o+1/n,l=t/n/2.506628274631)}return r>0?1-l:l}function b(t,n,e){return t<0||t>1?NaN:(n||0)+(null==e?1:e)*p*function(t){let n,e=-Math.log((1-t)*(1+t));e<6.25?(e-=3.125,n=-364441206401782e-35,n=n*e-16850591381820166e-35,n=128584807152564e-32+n*e,n=11157877678025181e-33+n*e,n=n*e-1333171662854621e-31,n=20972767875968562e-33+n*e,n=6637638134358324e-30+n*e,n=n*e-4054566272975207e-29,n=n*e-8151934197605472e-29,n=26335093153082323e-28+n*e,n=n*e-12975133253453532e-27,n=n*e-5415412054294628e-26,n=1.0512122733215323e-9+n*e,n=n*e-4.112633980346984e-9,n=n*e-2.9070369957882005e-8,n=4.2347877827932404e-7+n*e,n=n*e-13654692000834679e-22,n=n*e-13882523362786469e-21,n=.00018673420803405714+n*e,n=n*e-.000740702534166267,n=n*e-.006033670871430149,n=.24015818242558962+n*e,n=1.6536545626831027+n*e):e<16?(e=Math.sqrt(e)-3.25,n=2.2137376921775787e-9,n=9.075656193888539e-8+n*e,n=n*e-2.7517406297064545e-7,n=1.8239629214389228e-8+n*e,n=15027403968909828e-22+n*e,n=n*e-4013867526981546e-21,n=29234449089955446e-22+n*e,n=12475304481671779e-21+n*e,n=n*e-47318229009055734e-21,n=6828485145957318e-20+n*e,n=24031110387097894e-21+n*e,n=n*e-.0003550375203628475,n=.0009532893797373805+n*e,n=n*e-.0016882755560235047,n=.002491442096107851+n*e,n=n*e-.003751208507569241,n=.005370914553590064+n*e,n=1.0052589676941592+n*e,n=3.0838856104922208+n*e):Number.isFinite(e)?(e=Math.sqrt(e)-5,n=-27109920616438573e-27,n=n*e-2.555641816996525e-10,n=1.5076572693500548e-9+n*e,n=n*e-3.789465440126737e-9,n=7.61570120807834e-9+n*e,n=n*e-1.496002662714924e-8,n=2.914795345090108e-8+n*e,n=n*e-6.771199775845234e-8,n=2.2900482228026655e-7+n*e,n=n*e-9.9298272942317e-7,n=4526062597223154e-21+n*e,n=n*e-1968177810553167e-20,n=7599527703001776e-20+n*e,n=n*e-.00021503011930044477,n=n*e-.00013871931833623122,n=1.0103004648645344+n*e,n=4.849906401408584+n*e):n=1/0;return n*t}(2*t-1)}function v(t,n){let e,r;const o={mean(t){return arguments.length?(e=t||0,o):e},stdev(t){return arguments.length?(r=null==t?1:t,o):r},sample:()=>m(e,r),pdf:t=>g(t,e,r),cdf:t=>N(t,e,r),icdf:t=>b(t,e,r)};return o.mean(t).stdev(n)}function x(t,n){return t=t||0,n=null==n?1:n,Math.exp(t+m()*n)}function y(t,n,e){if(t<=0)return 0;n=n||0,e=null==e?1:e;const r=(Math.log(t)-n)/e;return Math.exp(-.5*r*r)/(e*d*t)}function w(t,n,e){return N(Math.log(t),n,e)}function q(t,n,e){return Math.exp(b(t,n,e))}function A(n,e){return null==e&&(e=null==n?1:n,n=0),n+(e-n)*t.random()}function F(t,n,e){return null==e&&(e=null==n?1:n,n=0),t>=n&&t<=e?1/(e-n):0}function L(t,n,e){return null==e&&(e=null==n?1:n,n=0),t<n?0:t>e?1:(t-n)/(e-n)}function S(t,n,e){return null==e&&(e=null==n?1:n,n=0),t>=0&&t<=1?n+t*(e-n):NaN}function E(t,n,e,r){const o=r-t*t,l=Math.abs(o)<1e-24?0:(e-t*n)/o;return[n-l*t,l]}function U(t,n,e,r){t=t.filter((t=>{let r=n(t),o=e(t);return null!=r&&(r=+r)>=r&&null!=o&&(o=+o)>=o})),r&&t.sort(((t,e)=>n(t)-n(e)));const o=t.length,l=new Float64Array(o),u=new Float64Array(o);let f,a,i,s=0,c=0,h=0;for(i of t)l[s]=f=+n(i),u[s]=a=+e(i),++s,c+=(f-c)/s,h+=(a-h)/s;for(s=0;s<o;++s)l[s]-=c,u[s]-=h;return[l,u,c,h]}function I(t,n,e,r){let o,l,u=-1;for(const f of t)o=n(f),l=e(f),null!=o&&(o=+o)>=o&&null!=l&&(l=+l)>=l&&r(o,l,++u)}function P(t,n,e,r,o){let l=0,u=0;return I(t,n,e,((t,n)=>{const e=n-o(t),f=n-r;l+=e*e,u+=f*f})),1-l/u}function T(t,n,e){let r=0,o=0,l=0,u=0,f=0;I(t,n,e,((t,n)=>{++f,r+=(t-r)/f,o+=(n-o)/f,l+=(t*n-l)/f,u+=(t*t-u)/f}));const a=E(r,o,l,u),i=t=>a[0]+a[1]*t;return{coef:a,predict:i,rSquared:P(t,n,e,o,i)}}function k(t,n,e){const[r,o,l,u]=U(t,n,e),f=r.length;let a,i,s,c,h=0,d=0,p=0,M=0,m=0;for(a=0;a<f;)i=r[a],s=o[a++],c=i*i,h+=(c-h)/a,d+=(c*i-d)/a,p+=(c*c-p)/a,M+=(i*s-M)/a,m+=(c*s-m)/a;const g=p-h*h,N=h*g-d*d,b=(m*h-M*d)/N,v=(M*g-m*d)/N,x=-b*h,y=t=>b*(t-=l)*t+v*t+x+u;return{coef:[x-v*l+b*l*l+u,v-2*b*l,b],predict:y,rSquared:P(t,n,e,u,y)}}function C(t,n,e,r){const o=Array(t);let l,u,f,a;for(l=0;l<t;++l)o[l]=0;for(l=t-1;l>=0;--l)for(f=n[l],a=1,o[l]+=f,u=1;u<=l;++u)a*=(l+1-u)/u,o[l-u]+=f*Math.pow(e,u)*a;return o[0]+=r,o}const D=2,R=1e-12;function K(t){return(t=1-t*t*t)*t*t}function Q(t,n,e){const r=t[n];let o=e[0],l=e[1]+1;if(!(l>=t.length))for(;n>o&&t[l]-r<=r-t[o];)e[0]=++o,e[1]=l,++l}const j=.5*Math.PI/180;function G(t,n,e,r,o){const l=Math.atan2(o*(e[1]-t[1]),r*(e[0]-t[0])),u=Math.atan2(o*(n[1]-t[1]),r*(n[0]-t[0]));return Math.abs(l-u)}t.bandwidthNRD=h,t.bin=function(t){const n=t.maxbins||20,e=t.base||10,r=Math.log(e),o=t.divide||[5,2];let l,u,f,a,i,s,c=t.extent[0],h=t.extent[1];const d=t.span||h-c||Math.abs(c)||1;if(t.step)l=t.step;else if(t.steps){for(a=d/n,i=0,s=t.steps.length;i<s&&t.steps[i]<a;++i);l=t.steps[Math.max(0,i-1)]}else{for(u=Math.ceil(Math.log(n)/r),f=t.minstep||0,l=Math.max(f,Math.pow(e,Math.round(Math.log(d)/r)-u));Math.ceil(d/l)>n;)l*=e;for(i=0,s=o.length;i<s;++i)a=l/o[i],a>=f&&d/a<=n&&(l=a)}a=Math.log(l);const p=a>=0?0:1+~~(-a/r),M=Math.pow(e,-p-1);return(t.nice||void 0===t.nice)&&(a=Math.floor(c/l+M)*l,c=c<a?a-l:a,h=Math.ceil(h/l)*l),{start:c,stop:h===c?c+l:h,step:l}},t.bootstrapCI=function(r,o,l,u){if(!r.length)return[void 0,void 0];const f=Float64Array.from(n(r,u)),a=f.length,s=o;let c,h,d,p;for(d=0,p=Array(s);d<s;++d){for(c=0,h=0;h<a;++h)c+=f[~~(t.random()*a)];p[d]=c/a}return p.sort(e),[i(p,l/2),i(p,1-l/2)]},t.cumulativeLogNormal=w,t.cumulativeNormal=N,t.cumulativeUniform=L,t.densityLogNormal=y,t.densityNormal=g,t.densityUniform=F,t.dotbin=function(t,n,e,r){r=r||(t=>t);const o=t.length,l=new Float64Array(o);let u,f=0,a=1,i=r(t[0]),s=i,c=i+n;for(;a<o;++a){if(u=r(t[a]),u>=c){for(s=(i+s)/2;f<a;++f)l[f]=s;c=u+n,i=u}s=u}for(s=(i+s)/2;f<a;++f)l[f]=s;return e?function(t,n){const e=t.length;let r,o,l=0,u=1;for(;t[l]===t[u];)++u;for(;u<e;){for(r=u+1;t[u]===t[r];)++r;if(t[u]-t[u-1]<n){for(o=u+(l+r-u-u>>1);o<u;)t[o++]=t[u];for(;o>u;)t[o--]=t[l]}l=u,u=r}return t}(l,n+n/4):l},t.quantileLogNormal=q,t.quantileNormal=b,t.quantileUniform=S,t.quantiles=s,t.quartiles=c,t.randomInteger=function(n,e){let r,o,l;null==e&&(e=n,n=0);const u={min(t){return arguments.length?(r=t||0,l=o-r,u):r},max(t){return arguments.length?(o=t||0,l=o-r,u):o},sample:()=>r+Math.floor(l*t.random()),pdf:t=>t===Math.floor(t)&&t>=r&&t<o?1/l:0,cdf(t){const n=Math.floor(t);return n<r?0:n>=o?1:(n-r+1)/l},icdf:t=>t>=0&&t<=1?r-1+Math.floor(t*l):NaN};return u.min(n).max(e)},t.randomKDE=function(n,e){const r=v();let o=0;const l={data(t){return arguments.length?(n=t,o=t?t.length:0,l.bandwidth(e)):n},bandwidth(t){return arguments.length?(!(e=t)&&n&&(e=h(n)),l):e},sample:()=>n[~~(t.random()*o)]+e*r.sample(),pdf(t){let l=0,u=0;for(;u<o;++u)l+=r.pdf((t-n[u])/e);return l/e/o},cdf(t){let l=0,u=0;for(;u<o;++u)l+=r.cdf((t-n[u])/e);return l/o},icdf(){throw Error("KDE icdf not supported.")}};return l.data(n)},t.randomLCG=function(t){return function(){return(t=(1103515245*t+12345)%2147483647)/2147483647}},t.randomLogNormal=function(t,n){let e,r;const o={mean(t){return arguments.length?(e=t||0,o):e},stdev(t){return arguments.length?(r=null==t?1:t,o):r},sample:()=>x(e,r),pdf:t=>y(t,e,r),cdf:t=>w(t,e,r),icdf:t=>q(t,e,r)};return o.mean(t).stdev(n)},t.randomMixture=function(n,e){let r,o=0;const l={weights(t){return arguments.length?(r=function(t){const n=[];let e,r=0;for(e=0;e<o;++e)r+=n[e]=null==t[e]?1:+t[e];for(e=0;e<o;++e)n[e]/=r;return n}(e=t||[]),l):e},distributions(t){return arguments.length?(t?(o=t.length,n=t):(o=0,n=[]),l.weights(e)):n},sample(){const e=t.random();let l=n[o-1],u=r[0],f=0;for(;f<o-1;u+=r[++f])if(e<u){l=n[f];break}return l.sample()},pdf(t){let e=0,l=0;for(;l<o;++l)e+=r[l]*n[l].pdf(t);return e},cdf(t){let e=0,l=0;for(;l<o;++l)e+=r[l]*n[l].cdf(t);return e},icdf(){throw Error("Mixture icdf not supported.")}};return l.distributions(n).weights(e)},t.randomNormal=v,t.randomUniform=function(t,n){let e,r;const o={min(t){return arguments.length?(e=t||0,o):e},max(t){return arguments.length?(r=null==t?1:t,o):r},sample:()=>A(e,r),pdf:t=>F(t,e,r),cdf:t=>L(t,e,r),icdf:t=>S(t,e,r)};return null==n&&(n=null==t?1:t,t=0),o.min(t).max(n)},t.regressionExp=function(t,n,e){const[r,o,l,u]=U(t,n,e);let f,a,i,s=0,c=0,h=0,d=0,p=0;I(t,n,e,((t,n)=>{f=r[p++],a=Math.log(n),i=f*n,s+=(n*a-s)/p,c+=(i-c)/p,h+=(i*a-h)/p,d+=(f*i-d)/p}));const[M,m]=E(c/u,s/u,h/u,d/u),g=t=>Math.exp(M+m*(t-l));return{coef:[Math.exp(M-m*l),m],predict:g,rSquared:P(t,n,e,u,g)}},t.regressionLinear=T,t.regressionLoess=function(t,n,e,r){const[o,l,u,f]=U(t,n,e,!0),a=o.length,s=Math.max(2,~~(r*a)),c=new Float64Array(a),h=new Float64Array(a),d=new Float64Array(a).fill(1);for(let t=-1;++t<=D;){const n=[0,s-1];for(let t=0;t<a;++t){const e=o[t],r=n[0],u=n[1],f=e-o[r]>o[u]-e?r:u;let a=0,i=0,s=0,p=0,M=0;const m=1/Math.abs(o[f]-e||1);for(let t=r;t<=u;++t){const n=o[t],r=l[t],u=K(Math.abs(e-n)*m)*d[t],f=n*u;a+=u,i+=f,s+=r*u,p+=r*f,M+=n*f}const[g,N]=E(i/a,s/a,p/a,M/a);c[t]=g+N*e,h[t]=Math.abs(l[t]-c[t]),Q(o,t+1,n)}if(t===D)break;const e=i(h,.5,void 0);if(Math.abs(e)<R)break;for(let t,n,r=0;r<a;++r)t=h[r]/(6*e),d[r]=t>=1?R:(n=1-t*t)*n}return function(t,n,e,r){const o=t.length,l=[];let u,f=0,a=0,i=[];for(;f<o;++f)u=t[f]+e,i[0]===u?i[1]+=(n[f]-i[1])/++a:(a=0,i[1]+=r,i=[u,n[f]],l.push(i));return i[1]+=r,l}(o,c,u,f)},t.regressionLog=function(t,n,e){let r=0,o=0,l=0,u=0,f=0;I(t,n,e,((t,n)=>{++f,t=Math.log(t),r+=(t-r)/f,o+=(n-o)/f,l+=(t*n-l)/f,u+=(t*t-u)/f}));const a=E(r,o,l,u),i=t=>a[0]+a[1]*Math.log(t);return{coef:a,predict:i,rSquared:P(t,n,e,o,i)}},t.regressionPoly=function(t,n,e,r){if(1===r)return T(t,n,e);if(2===r)return k(t,n,e);const[o,l,u,f]=U(t,n,e),a=o.length,i=[],s=[],c=r+1;let h,d,p,M,m;for(h=0;h<c;++h){for(p=0,M=0;p<a;++p)M+=Math.pow(o[p],h)*l[p];for(i.push(M),m=new Float64Array(c),d=0;d<c;++d){for(p=0,M=0;p<a;++p)M+=Math.pow(o[p],h+d);m[d]=M}s.push(m)}s.push(i);const g=function(t){const n=t.length-1,e=[];let r,o,l,u,f;for(r=0;r<n;++r){for(u=r,o=r+1;o<n;++o)Math.abs(t[r][o])>Math.abs(t[r][u])&&(u=o);for(l=r;l<n+1;++l)f=t[l][r],t[l][r]=t[l][u],t[l][u]=f;for(o=r+1;o<n;++o)for(l=n;l>=r;l--)t[l][o]-=t[l][r]*t[r][o]/t[r][r]}for(o=n-1;o>=0;--o){for(f=0,l=o+1;l<n;++l)f+=t[l][o]*e[l];e[o]=(t[n][o]-f)/t[o][o]}return e}(s),N=t=>{t-=u;let n=f+g[0]+g[1]*t+g[2]*t*t;for(h=3;h<c;++h)n+=g[h]*Math.pow(t,h);return n};return{coef:C(c,g,-u,f),predict:N,rSquared:P(t,n,e,f,N)}},t.regressionPow=function(t,n,e){let r=0,o=0,l=0,u=0,f=0,a=0;I(t,n,e,((t,n)=>{const e=Math.log(t),i=Math.log(n);++a,r+=(e-r)/a,o+=(i-o)/a,l+=(e*i-l)/a,u+=(e*e-u)/a,f+=(n-f)/a}));const i=E(r,o,l,u),s=t=>i[0]*Math.pow(t,i[1]);return i[0]=Math.exp(i[0]),{coef:i,predict:s,rSquared:P(t,n,e,f,s)}},t.regressionQuad=k,t.sampleCurve=function(t,n,e,r){e=e||25,r=Math.max(e,r||200);const o=n=>[n,t(n)],l=n[0],u=n[1],f=u-l,a=f/r,i=[o(l)],s=[];if(e===r){for(let t=1;t<r;++t)i.push(o(l+t/e*f));return i.push(o(u)),i}s.push(o(u));for(let t=e;--t>0;)s.push(o(l+t/e*f));let c=i[0],h=s[s.length-1];const d=1/f,p=function(t,n){let e=t,r=t;const o=n.length;for(let t=0;t<o;++t){const o=n[t][1];o<e&&(e=o),o>r&&(r=o)}return 1/(r-e)}(c[1],s);for(;h;){const t=o((c[0]+h[0])/2);t[0]-c[0]>=a&&G(c,t,h,d,p)>j?s.push(t):(c=h,i.push(h),s.pop()),h=s[s.length-1]}return i},t.sampleLogNormal=x,t.sampleNormal=m,t.sampleUniform=A,t.setRandom=function(n){t.random=n}}));
//# sourceMappingURL=vega-statistics.min.js.map

527

build/vega-statistics.module.js

@@ -12,6 +12,4 @@ import { ascending, quantileSorted, deviation, quantile, median } from 'd3-array';

let index = -1;
for (let value of values) {
value = valueof(value, ++index, values);
if (value != null && value !== '' && (value = +value) >= value) {

@@ -25,5 +23,6 @@ yield value;

function quantiles (array, p, f) {
const values = Float64Array.from(numbers(array, f)); // don't depend on return value from typed array sort call
const values = Float64Array.from(numbers(array, f));
// don't depend on return value from typed array sort call
// protects against undefined sort results in Safari (vega/vega-lite#4964)
values.sort(ascending);

@@ -37,10 +36,10 @@ return p.map(_ => quantileSorted(values, _));

// Scott, D. W. (1992) Multivariate Density Estimation:
// Theory, Practice, and Visualization. Wiley.
function estimateBandwidth (array, f) {
const n = array.length,
d = deviation(array, f),
q = quartiles(array, f),
h = (q[2] - q[0]) / 1.34,
v = Math.min(d, h) || d || Math.abs(q[0]) || 1;
d = deviation(array, f),
q = quartiles(array, f),
h = (q[2] - q[0]) / 1.34,
v = Math.min(d, h) || d || Math.abs(q[0]) || 1;
return 1.06 * v * Math.pow(n, -0.2);

@@ -52,15 +51,14 @@ }

const maxb = _.maxbins || 20,
base = _.base || 10,
logb = Math.log(base),
div = _.divide || [5, 2];
base = _.base || 10,
logb = Math.log(base),
div = _.divide || [5, 2];
let min = _.extent[0],
max = _.extent[1],
step,
level,
minstep,
v,
i,
n;
max = _.extent[1],
step,
level,
minstep,
v,
i,
n;
const span = _.span || max - min || Math.abs(min) || 1;
if (_.step) {

@@ -72,5 +70,3 @@ // if step size is explicitly given, use that

v = span / maxb;
for (i = 0, n = _.steps.length; i < n && _.steps[i] < v; ++i);
step = _.steps[Math.max(0, i - 1)];

@@ -81,9 +77,10 @@ } else {

minstep = _.minstep || 0;
step = Math.max(minstep, Math.pow(base, Math.round(Math.log(span) / logb) - level)); // increase step size if too many bins
step = Math.max(minstep, Math.pow(base, Math.round(Math.log(span) / logb) - level));
// increase step size if too many bins
while (Math.ceil(span / step) > maxb) {
step *= base;
} // decrease step size if allowed
}
// decrease step size if allowed
for (i = 0, n = div.length; i < n; ++i) {

@@ -93,9 +90,8 @@ v = step / div[i];

}
} // update precision, min and max
}
// update precision, min and max
v = Math.log(step);
const precision = v >= 0 ? 0 : ~~(-v / logb) + 1,
eps = Math.pow(base, -precision - 1);
eps = Math.pow(base, -precision - 1);
if (_.nice || _.nice === undefined) {

@@ -106,3 +102,2 @@ v = Math.floor(min / step + eps) * step;

}
return {

@@ -123,6 +118,5 @@ start: min,

const values = Float64Array.from(numbers(array, f)),
n = values.length,
m = samples;
n = values.length,
m = samples;
let a, i, j, mu;
for (j = 0, mu = Array(m); j < m; ++j) {

@@ -132,6 +126,4 @@ for (a = 0, i = 0; i < n; ++i) {

}
mu[j] = a / n;
}
mu.sort(ascending);

@@ -146,66 +138,54 @@ return [quantile(mu, alpha / 2), quantile(mu, 1 - alpha / 2)];

f = f || (_ => _);
const n = array.length,
v = new Float64Array(n);
v = new Float64Array(n);
let i = 0,
j = 1,
a = f(array[0]),
b = a,
w = a + step,
x;
j = 1,
a = f(array[0]),
b = a,
w = a + step,
x;
for (; j < n; ++j) {
x = f(array[j]);
if (x >= w) {
b = (a + b) / 2;
for (; i < j; ++i) v[i] = b;
w = x + step;
a = x;
}
b = x;
}
b = (a + b) / 2;
for (; i < j; ++i) v[i] = b;
return smooth ? smoothing(v, step + step / 4) : v;
}
return smooth ? smoothing(v, step + step / 4) : v;
} // perform smoothing to reduce variance
// perform smoothing to reduce variance
// swap points between "adjacent" stacks
// Wilkinson defines adjacent as within step/4 units
function smoothing(v, thresh) {
const n = v.length;
let a = 0,
b = 1,
c,
d; // get left stack
b = 1,
c,
d;
// get left stack
while (v[a] === v[b]) ++b;
while (b < n) {
// get right stack
c = b + 1;
while (v[b] === v[c]) ++c;
while (v[b] === v[c]) ++c; // are stacks adjacent?
// are stacks adjacent?
// if so, compare sizes and swap as needed
if (v[b] - v[b - 1] < thresh) {
d = b + (a + c - b - b >> 1);
while (d < b) v[d++] = v[b];
while (d > b) v[d--] = v[a];
} // update left stack indices
}
// update left stack indices
a = b;
b = c;
}
return v;

@@ -228,3 +208,2 @@ }

}
let a, b, d;

@@ -241,3 +220,2 @@ const dist = {

},
max(_) {

@@ -252,11 +230,8 @@ if (arguments.length) {

},
sample() {
return a + Math.floor(d * random());
},
pdf(x) {
return x === Math.floor(x) && x >= a && x < b ? 1 / d : 0;
},
cdf(x) {

@@ -266,7 +241,5 @@ const v = Math.floor(x);

},
icdf(p) {
return p >= 0 && p <= 1 ? a - 1 + Math.floor(p * d) : NaN;
}
};

@@ -284,6 +257,5 @@ return dist.min(min).max(max);

let x = 0,
y = 0,
rds,
c;
y = 0,
rds,
c;
if (nextSample === nextSample) {

@@ -298,9 +270,6 @@ x = nextSample;

} while (rds === 0 || rds > 1);
c = Math.sqrt(-2 * Math.log(rds) / rds); // Box-Muller transform
x *= c;
nextSample = y * c;
}
return mean + x * stdev;

@@ -312,5 +281,6 @@ }

return Math.exp(-0.5 * z * z) / (stdev * SQRT2PI);
} // Approximation from West (2009)
}
// Approximation from West (2009)
// Better Approximations to Cumulative Normal Functions
function cumulativeNormal(value, mean, stdev) {

@@ -320,5 +290,4 @@ mean = mean || 0;

const z = (value - mean) / stdev,
Z = Math.abs(z);
Z = Math.abs(z);
let cd;
if (Z > 37) {

@@ -329,3 +298,2 @@ cd = 0;

let sum;
if (Z < 7.07106781186547) {

@@ -356,13 +324,14 @@ sum = 3.52624965998911e-02 * Z + 0.700383064443688;

}
return z > 0 ? 1 - cd : cd;
} // Approximation of Probit function using inverse error function.
}
// Approximation of Probit function using inverse error function.
function quantileNormal(p, mean, stdev) {
if (p < 0 || p > 1) return NaN;
return (mean || 0) + (stdev == null ? 1 : stdev) * SQRT2 * erfinv(2 * p - 1);
} // Approximate inverse error function. Implementation from "Approximating
}
// Approximate inverse error function. Implementation from "Approximating
// the erfinv function" by Mike Giles, GPU Computing Gems, volume 2, 2010.
// Ported from Apache Commons Math, http://www.apache.org/licenses/LICENSE-2.0
function erfinv(x) {

@@ -374,4 +343,3 @@ // beware that the logarithm argument must be

let w = -Math.log((1 - x) * (1 + x)),
p;
p;
if (w < 6.25) {

@@ -445,6 +413,4 @@ w -= 3.125;

}
return p * x;
}
function gaussian (mean, stdev) {

@@ -461,3 +427,2 @@ let mu, sigma;

},
stdev(_) {

@@ -471,3 +436,2 @@ if (arguments.length) {

},
sample: () => sampleNormal(mu, sigma),

@@ -494,3 +458,2 @@ pdf: value => densityNormal(value, mu, sigma),

},
bandwidth(_) {

@@ -502,33 +465,24 @@ if (!arguments.length) return bandwidth;

},
sample() {
return support[~~(random() * n)] + bandwidth * kernel.sample();
},
pdf(x) {
let y = 0,
i = 0;
i = 0;
for (; i < n; ++i) {
y += kernel.pdf((x - support[i]) / bandwidth);
}
return y / bandwidth / n;
},
cdf(x) {
let y = 0,
i = 0;
i = 0;
for (; i < n; ++i) {
y += kernel.cdf((x - support[i]) / bandwidth);
}
return y / n;
},
icdf() {
throw Error('KDE icdf not supported.');
}
};

@@ -567,3 +521,2 @@ return dist.data(support);

},
stdev(_) {

@@ -577,3 +530,2 @@ if (arguments.length) {

},
sample: () => sampleLogNormal(mu, sigma),

@@ -589,20 +541,15 @@ pdf: value => densityLogNormal(value, mu, sigma),

let m = 0,
w;
w;
function normalize(x) {
const w = [];
let sum = 0,
i;
i;
for (i = 0; i < m; ++i) {
sum += w[i] = x[i] == null ? 1 : +x[i];
}
for (i = 0; i < m; ++i) {
w[i] /= sum;
}
return w;
}
const dist = {

@@ -614,6 +561,4 @@ weights(_) {

}
return weights;
},
distributions(_) {

@@ -628,15 +573,13 @@ if (arguments.length) {

}
return dist.weights(weights);
}
return dists;
},
sample() {
const r = random();
let d = dists[m - 1],
v = w[0],
i = 0; // first select distribution
v = w[0],
i = 0;
// first select distribution
for (; i < m - 1; v += w[++i]) {

@@ -647,34 +590,25 @@ if (r < v) {

}
} // then sample from it
}
// then sample from it
return d.sample();
},
pdf(x) {
let p = 0,
i = 0;
i = 0;
for (; i < m; ++i) {
p += w[i] * dists[i].pdf(x);
}
return p;
},
cdf(x) {
let p = 0,
i = 0;
i = 0;
for (; i < m; ++i) {
p += w[i] * dists[i].cdf(x);
}
return p;
},
icdf() {
throw Error('Mixture icdf not supported.');
}
};

@@ -689,3 +623,2 @@ return dist.distributions(dists).weights(weights);

}
return min + (max - min) * random();

@@ -698,3 +631,2 @@ }

}
return value >= min && value <= max ? 1 / (max - min) : 0;

@@ -707,3 +639,2 @@ }

}
return value < min ? 0 : value > max ? 1 : (value - min) / (max - min);

@@ -716,3 +647,2 @@ }

}
return p >= 0 && p <= 1 ? min + p * (max - min) : NaN;

@@ -731,3 +661,2 @@ }

},
max(_) {

@@ -741,3 +670,2 @@ if (arguments.length) {

},
sample: () => sampleUniform(a, b),

@@ -748,3 +676,2 @@ pdf: value => densityUniform(value, a, b),

};
if (max == null) {

@@ -754,3 +681,2 @@ max = min == null ? 1 : min;

}
return dist.min(min).max(max);

@@ -762,4 +688,4 @@ }

const delta = uX2 - uX * uX,
slope = Math.abs(delta) < 1e-24 ? 0 : (uXY - uX * uY) / delta,
intercept = uY - slope * uX;
slope = Math.abs(delta) < 1e-24 ? 0 : (uXY - uX * uY) / delta,
intercept = uY - slope * uX;
return [intercept, slope];

@@ -771,21 +697,19 @@ }

let u = x(d),
v = y(d);
v = y(d);
return u != null && (u = +u) >= u && v != null && (v = +v) >= v;
});
if (sort) {
data.sort((a, b) => x(a) - x(b));
}
const n = data.length,
X = new Float64Array(n),
Y = new Float64Array(n); // extract values, calculate means
X = new Float64Array(n),
Y = new Float64Array(n);
// extract values, calculate means
let i = 0,
ux = 0,
uy = 0,
xv,
yv,
d;
ux = 0,
uy = 0,
xv,
yv,
d;
for (d of data) {

@@ -797,5 +721,5 @@ X[i] = xv = +x(d);

uy += (yv - uy) / i;
} // mean center the data
}
// mean center the data
for (i = 0; i < n; ++i) {

@@ -805,3 +729,2 @@ X[i] -= ux;

}
return [X, Y, ux, uy];

@@ -811,9 +734,7 @@ }

let i = -1,
u,
v;
u,
v;
for (const d of data) {
u = x(d);
v = y(d);
if (u != null && (u = +u) >= u && v != null && (v = +v) >= v) {

@@ -825,10 +746,10 @@ callback(u, v, ++i);

// Adapted from d3-regression by Harry Stevens
// License: https://github.com/HarryStevens/d3-regression/blob/master/LICENSE
function rSquared (data, x, y, uY, predict) {
let SSE = 0,
SST = 0;
SST = 0;
visitPoints(data, x, y, (dx, dy) => {
const sse = dy - predict(dx),
sst = dy - uY;
sst = dy - uY;
SSE += sse * sse;

@@ -840,10 +761,10 @@ SST += sst * sst;

// Adapted from d3-regression by Harry Stevens
// License: https://github.com/HarryStevens/d3-regression/blob/master/LICENSE
function linear (data, x, y) {
let X = 0,
Y = 0,
XY = 0,
X2 = 0,
n = 0;
Y = 0,
XY = 0,
X2 = 0,
n = 0;
visitPoints(data, x, y, (dx, dy) => {

@@ -856,6 +777,4 @@ ++n;

});
const coef = ols(X, Y, XY, X2),
predict = x => coef[0] + coef[1] * x;
predict = x => coef[0] + coef[1] * x;
return {

@@ -868,10 +787,10 @@ coef: coef,

// Adapted from d3-regression by Harry Stevens
// License: https://github.com/HarryStevens/d3-regression/blob/master/LICENSE
function log (data, x, y) {
let X = 0,
Y = 0,
XY = 0,
X2 = 0,
n = 0;
Y = 0,
XY = 0,
X2 = 0,
n = 0;
visitPoints(data, x, y, (dx, dy) => {

@@ -885,6 +804,4 @@ ++n;

});
const coef = ols(X, Y, XY, X2),
predict = x => coef[0] + coef[1] * Math.log(x);
predict = x => coef[0] + coef[1] * Math.log(x);
return {

@@ -901,9 +818,9 @@ coef: coef,

let YL = 0,
XY = 0,
XYL = 0,
X2Y = 0,
n = 0,
dx,
ly,
xy;
XY = 0,
XYL = 0,
X2Y = 0,
n = 0,
dx,
ly,
xy;
visitPoints(data, x, y, (_, dy) => {

@@ -918,6 +835,4 @@ dx = xv[n++];

});
const [c0, c1] = ols(XY / uy, YL / uy, XYL / uy, X2Y / uy),
predict = x => Math.exp(c0 + c1 * (x - ux));
predict = x => Math.exp(c0 + c1 * (x - ux));
return {

@@ -930,14 +845,14 @@ coef: [Math.exp(c0 - c1 * ux), c1],

// Adapted from d3-regression by Harry Stevens
// License: https://github.com/HarryStevens/d3-regression/blob/master/LICENSE
function pow (data, x, y) {
let X = 0,
Y = 0,
XY = 0,
X2 = 0,
YS = 0,
n = 0;
Y = 0,
XY = 0,
X2 = 0,
YS = 0,
n = 0;
visitPoints(data, x, y, (dx, dy) => {
const lx = Math.log(dx),
ly = Math.log(dy);
ly = Math.log(dy);
++n;

@@ -950,6 +865,4 @@ X += (lx - X) / n;

});
const coef = ols(X, Y, XY, X2),
predict = x => coef[0] * Math.pow(x, coef[1]);
predict = x => coef[0] * Math.pow(x, coef[1]);
coef[0] = Math.exp(coef[0]);

@@ -965,13 +878,12 @@ return {

const [xv, yv, ux, uy] = points(data, x, y),
n = xv.length;
n = xv.length;
let X2 = 0,
X3 = 0,
X4 = 0,
XY = 0,
X2Y = 0,
i,
dx,
dy,
x2;
X3 = 0,
X4 = 0,
XY = 0,
X2Y = 0,
i,
dx,
dy,
x2;
for (i = 0; i < n;) {

@@ -987,14 +899,13 @@ dx = xv[i];

}
const X2X2 = X4 - X2 * X2,
d = X2 * X2X2 - X3 * X3,
a = (X2Y * X2 - XY * X3) / d,
b = (XY * X2X2 - X2Y * X3) / d,
c = -a * X2,
predict = x => {
x = x - ux;
return a * x * x + b * x + c + uy;
}; // transform coefficients back from mean-centered space
d = X2 * X2X2 - X3 * X3,
a = (X2Y * X2 - XY * X3) / d,
b = (XY * X2X2 - X2Y * X3) / d,
c = -a * X2,
predict = x => {
x = x - ux;
return a * x * x + b * x + c + uy;
};
// transform coefficients back from mean-centered space
return {

@@ -1007,2 +918,3 @@ coef: [c - b * ux + a * ux * ux + uy, b - 2 * a * ux, a],

// Adapted from d3-regression by Harry Stevens
// License: https://github.com/HarryStevens/d3-regression/blob/master/LICENSE

@@ -1012,3 +924,2 @@ // ... which was adapted from regression-js by Tom Alexander

// License: https://github.com/Tom-Alexander/regression-js/blob/master/LICENSE
function poly (data, x, y, order) {

@@ -1019,8 +930,7 @@ // use more efficient methods for lower orders

const [xv, yv, ux, uy] = points(data, x, y),
n = xv.length,
lhs = [],
rhs = [],
k = order + 1;
n = xv.length,
lhs = [],
rhs = [],
k = order + 1;
let i, j, l, v, c;
for (i = 0; i < k; ++i) {

@@ -1030,6 +940,4 @@ for (l = 0, v = 0; l < n; ++l) {

}
lhs.push(v);
c = new Float64Array(k);
for (j = 0; j < k; ++j) {

@@ -1039,21 +947,14 @@ for (l = 0, v = 0; l < n; ++l) {

}
c[j] = v;
}
rhs.push(c);
}
rhs.push(lhs);
const coef = gaussianElimination(rhs),
predict = x => {
x -= ux;
let y = uy + coef[0] + coef[1] * x + coef[2] * x * x;
for (i = 3; i < k; ++i) y += coef[i] * Math.pow(x, i);
return y;
};
predict = x => {
x -= ux;
let y = uy + coef[0] + coef[1] * x + coef[2] * x * x;
for (i = 3; i < k; ++i) y += coef[i] * Math.pow(x, i);
return y;
};
return {

@@ -1065,10 +966,10 @@ coef: uncenter(k, coef, -ux, uy),

}
function uncenter(k, a, x, y) {
const z = Array(k);
let i, j, v, c; // initialize to zero
let i, j, v, c;
for (i = 0; i < k; ++i) z[i] = 0; // polynomial expansion
// initialize to zero
for (i = 0; i < k; ++i) z[i] = 0;
// polynomial expansion
for (i = k - 1; i >= 0; --i) {

@@ -1078,25 +979,21 @@ v = a[i];

z[i] += v;
for (j = 1; j <= i; ++j) {
c *= (i + 1 - j) / j; // binomial coefficent
z[i - j] += v * Math.pow(x, j) * c;
}
} // bias term
}
// bias term
z[0] += y;
return z;
} // Given an array for a two-dimensional matrix and the polynomial order,
}
// Given an array for a two-dimensional matrix and the polynomial order,
// solve A * x = b using Gaussian elimination.
function gaussianElimination(matrix) {
const n = matrix.length - 1,
coef = [];
coef = [];
let i, j, k, r, t;
for (i = 0; i < n; ++i) {
r = i; // max row
for (j = i + 1; j < n; ++j) {

@@ -1107,3 +1004,2 @@ if (Math.abs(matrix[i][j]) > Math.abs(matrix[i][r])) {

}
for (k = i; k < n + 1; ++k) {

@@ -1114,3 +1010,2 @@ t = matrix[k][i];

}
for (j = i + 1; j < n; ++j) {

@@ -1122,13 +1017,9 @@ for (k = n; k >= i; k--) {

}
for (j = n - 1; j >= 0; --j) {
t = 0;
for (k = j + 1; k < n; ++k) {
t += matrix[k][j] * coef[k];
}
coef[j] = (matrix[n][j] - t) / matrix[j][j];
}
return coef;

@@ -1138,28 +1029,27 @@ }

const maxiters = 2,
epsilon = 1e-12; // Adapted from science.js by Jason Davies
epsilon = 1e-12;
// Adapted from science.js by Jason Davies
// Source: https://github.com/jasondavies/science.js/blob/master/src/stats/loess.js
// License: https://github.com/jasondavies/science.js/blob/master/LICENSE
function loess (data, x, y, bandwidth) {
const [xv, yv, ux, uy] = points(data, x, y, true),
n = xv.length,
bw = Math.max(2, ~~(bandwidth * n)),
// # nearest neighbors
yhat = new Float64Array(n),
residuals = new Float64Array(n),
robustWeights = new Float64Array(n).fill(1);
n = xv.length,
bw = Math.max(2, ~~(bandwidth * n)),
// # nearest neighbors
yhat = new Float64Array(n),
residuals = new Float64Array(n),
robustWeights = new Float64Array(n).fill(1);
for (let iter = -1; ++iter <= maxiters;) {
const interval = [0, bw - 1];
for (let i = 0; i < n; ++i) {
const dx = xv[i],
i0 = interval[0],
i1 = interval[1],
edge = dx - xv[i0] > xv[i1] - dx ? i0 : i1;
i0 = interval[0],
i1 = interval[1],
edge = dx - xv[i0] > xv[i1] - dx ? i0 : i1;
let W = 0,
X = 0,
Y = 0,
XY = 0,
X2 = 0;
X = 0,
Y = 0,
XY = 0,
X2 = 0;
const denom = 1 / Math.abs(xv[edge] - dx || 1); // avoid singularity!

@@ -1169,5 +1059,5 @@

const xk = xv[k],
yk = yv[k],
w = tricube(Math.abs(dx - xk) * denom) * robustWeights[k],
xkw = xk * w;
yk = yv[k],
w = tricube(Math.abs(dx - xk) * denom) * robustWeights[k],
xkw = xk * w;
W += w;

@@ -1178,5 +1068,5 @@ X += xkw;

X2 += xk * xkw;
} // linear regression fit
}
// linear regression fit
const [a, b] = ols(X / W, Y / W, XY / W, X2 / W);

@@ -1187,33 +1077,31 @@ yhat[i] = a + b * dx;

}
if (iter === maxiters) {
break;
}
const medianResidual = median(residuals);
if (Math.abs(medianResidual) < epsilon) break;
for (let i = 0, arg, w; i < n; ++i) {
arg = residuals[i] / (6 * medianResidual); // default to epsilon (rather than zero) for large deviations
arg = residuals[i] / (6 * medianResidual);
// default to epsilon (rather than zero) for large deviations
// keeping weights tiny but non-zero prevents singularites
robustWeights[i] = arg >= 1 ? epsilon : (w = 1 - arg * arg) * w;
}
}
return output(xv, yhat, ux, uy);
} // weighting kernel for local regression
}
// weighting kernel for local regression
function tricube(x) {
return (x = 1 - x * x * x) * x * x;
} // advance sliding window interval of nearest neighbors
}
// advance sliding window interval of nearest neighbors
function updateInterval(xv, i, interval) {
const val = xv[i];
let left = interval[0],
right = interval[1] + 1;
if (right >= xv.length) return; // step right if distance to new right edge is <= distance to old left edge
right = interval[1] + 1;
if (right >= xv.length) return;
// step right if distance to new right edge is <= distance to old left edge
// step when distance is equal to ensure movement over duplicate x values
while (i > left && xv[right] - val <= val - xv[left]) {

@@ -1224,17 +1112,15 @@ interval[0] = ++left;

}
} // generate smoothed output points
}
// generate smoothed output points
// average points with repeated x values
function output(xv, yhat, ux, uy) {
const n = xv.length,
out = [];
out = [];
let i = 0,
cnt = 0,
prev = [],
v;
cnt = 0,
prev = [],
v;
for (; i < n; ++i) {
v = xv[i] + ux;
if (prev[0] === v) {

@@ -1251,3 +1137,2 @@ // average output values via online update

}
prev[1] += uy;

@@ -1258,16 +1143,15 @@ return out;

// subdivide up to accuracy of 0.5 degrees
const MIN_RADIANS = 0.5 * Math.PI / 180; // Adaptively sample an interpolated function over a domain extent
const MIN_RADIANS = 0.5 * Math.PI / 180;
// Adaptively sample an interpolated function over a domain extent
function sampleCurve (f, extent, minSteps, maxSteps) {
minSteps = minSteps || 25;
maxSteps = Math.max(minSteps, maxSteps || 200);
const point = x => [x, f(x)],
minX = extent[0],
maxX = extent[1],
span = maxX - minX,
stop = span / maxSteps,
prev = [point(minX)],
next = [];
minX = extent[0],
maxX = extent[1],
span = maxX - minX,
stop = span / maxSteps,
prev = [point(minX)],
next = [];
if (minSteps === maxSteps) {

@@ -1278,3 +1162,2 @@ // no adaptation, sample uniform grid directly and return

}
prev.push(point(maxX));

@@ -1286,3 +1169,2 @@ return prev;

next.push(point(maxX));
for (let i = minSteps; --i > 0;) {

@@ -1292,3 +1174,2 @@ next.push(point(minX + i / minSteps * span));

}
let p0 = prev[0];

@@ -1298,3 +1179,2 @@ let p1 = next[next.length - 1];

const sy = scaleY(p0[1], next);
while (p1) {

@@ -1304,3 +1184,2 @@ // midpoint for potential curve subdivision

const dx = pm[0] - p0[0] >= stop;
if (dx && angleDelta(p0, pm, p1, sx, sy) > MIN_RADIANS) {

@@ -1318,9 +1197,6 @@ // maximum resolution has not yet been met, and

}
p1 = next[next.length - 1];
}
return prev;
}
function scaleY(init, points) {

@@ -1330,3 +1206,2 @@ let ymin = init;

const n = points.length;
for (let i = 0; i < n; ++i) {

@@ -1337,9 +1212,7 @@ const y = points[i][1];

}
return 1 / (ymax - ymin);
}
function angleDelta(p, q, r, sx, sy) {
const a0 = Math.atan2(sy * (r[1] - p[1]), sx * (r[0] - p[0])),
a1 = Math.atan2(sy * (q[1] - p[1]), sx * (q[0] - p[0]));
a1 = Math.atan2(sy * (q[1] - p[1]), sx * (q[0] - p[0]));
return Math.abs(a0 - a1);

@@ -1346,0 +1219,0 @@ }

8

package.json
{
"name": "vega-statistics",
"version": "1.8.0",
"version": "1.8.1",
"description": "Statistical routines and probability distributions.",

@@ -19,3 +19,3 @@ "keywords": [

"prebuild": "rimraf build",
"build": "rollup -c",
"build": "rollup -c rollup.config.mjs",
"pretest": "yarn build --config-test",

@@ -26,5 +26,5 @@ "test": "tape 'test/**/*-test.js'",

"dependencies": {
"d3-array": "^3.1.1"
"d3-array": "^3.2.2"
},
"gitHead": "9a3faca4395cade9ecdfde90af98f1c53e9916b2"
"gitHead": "fb1092f6b931d450f9c210b67ae4752bd3dd461b"
}
rollup.config.js
build/vega-statistics.min.js.map

Sorry, the diff of this file is not supported yet

LICENSE

Sorry, the diff of this file is not supported yet

SocketSocket SOC 2 Logo

Product

  • Package Alerts
  • Integrations
  • Docs
  • Pricing
  • FAQ
  • Roadmap
  • Changelog

About

Packages

npm

Go

Maven

PyPI

Rubygems

Stay in touch

Get open source security insights delivered straight into your inbox.

  • Terms
  • Privacy
  • Security

Made with ⚡️ by Socket Inc