267 lines
7.1 KiB
JavaScript
267 lines
7.1 KiB
JavaScript
/**
|
|
* @license Apache-2.0
|
|
*
|
|
* Copyright (c) 2018 The Stdlib Authors.
|
|
*
|
|
* Licensed under the Apache License, Version 2.0 (the "License");
|
|
* you may not use this file except in compliance with the License.
|
|
* You may obtain a copy of the License at
|
|
*
|
|
* http://www.apache.org/licenses/LICENSE-2.0
|
|
*
|
|
* Unless required by applicable law or agreed to in writing, software
|
|
* distributed under the License is distributed on an "AS IS" BASIS,
|
|
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
|
* See the License for the specific language governing permissions and
|
|
* limitations under the License.
|
|
*/
|
|
|
|
'use strict';
|
|
|
|
// MODULES //
|
|
|
|
var isPositiveInteger = require( '@stdlib/assert/is-positive-integer' ).isPrimitive;
|
|
var isArrayLike = require( '@stdlib/assert/is-array-like-object' );
|
|
var isnan = require( '@stdlib/math/base/assert/is-nan' );
|
|
var Float64Array = require( '@stdlib/array/float64' );
|
|
|
|
|
|
// MAIN //
|
|
|
|
/**
|
|
* Returns an accumulator function which incrementally computes a moving arithmetic mean and unbiased sample variance.
|
|
*
|
|
* ## Method
|
|
*
|
|
* - Let \\(W\\) be a window of \\(N\\) elements over which we want to compute an unbiased sample variance.
|
|
*
|
|
* - The difference between the unbiased sample variance in a window \\(W_i\\) and the unbiased sample variance in a window \\(W_{i+1})\\) is given by
|
|
*
|
|
* ```tex
|
|
* \Delta s^2 = s_{i+1}^2 - s_{i}^2
|
|
* ```
|
|
*
|
|
* - If we multiply both sides by \\(N-1\\),
|
|
*
|
|
* ```tex
|
|
* (N-1)(\Delta s^2) = (N-1)s_{i+1}^2 - (N-1)s_{i}^2
|
|
* ```
|
|
*
|
|
* - If we substitute the definition of the unbiased sample variance having the form
|
|
*
|
|
* ```tex
|
|
* \begin{align*}
|
|
* s^2 &= \frac{1}{N-1} \biggl( \sum_{i=1}^{N} (x_i - \bar{x})^2 \biggr) \\
|
|
* &= \frac{1}{N-1} \biggl( \sum_{i=1}^{N} (x_i^2 - 2\bar{x}x_i + \bar{x}^2) \biggr) \\
|
|
* &= \frac{1}{N-1} \biggl( \sum_{i=1}^{N} x_i^2 - 2\bar{x} \sum_{i=1}^{N} x_i + \sum_{i=1}^{N} \bar{x}^2) \biggr) \\
|
|
* &= \frac{1}{N-1} \biggl( \sum_{i=1}^{N} x_i^2 - \frac{2N\bar{x}\sum_{i=1}^{N} x_i}{N} + N\bar{x}^2 \biggr) \\
|
|
* &= \frac{1}{N-1} \biggl( \sum_{i=1}^{N} x_i^2 - 2N\bar{x}^2 + N\bar{x}^2 \biggr) \\
|
|
* &= \frac{1}{N-1} \biggl( \sum_{i=1}^{N} x_i^2 - N\bar{x}^2 \biggr)
|
|
* \end{align*}
|
|
* ```
|
|
*
|
|
* we return
|
|
*
|
|
* ```tex
|
|
* (N-1)(\Delta s^2) = \biggl(\sum_{k=1}^N x_k^2 - N\bar{x}_{i+1}^2 \biggr) - \biggl(\sum_{k=0}^{N-1} x_k^2 - N\bar{x}_{i}^2 \biggr)
|
|
* ```
|
|
*
|
|
* - This can be further simplified by recognizing that subtracting the sums reduces to \\(x_N^2 - x_0^2\\); in which case,
|
|
*
|
|
* ```tex
|
|
* \begin{align*}
|
|
* (N-1)(\Delta s^2) &= x_N^2 - x_0^2 - N\bar{x}_{i+1}^2 + N\bar{x}_{i}^2 \\
|
|
* &= x_N^2 - x_0^2 - N(\bar{x}_{i+1}^2 - \bar{x}_{i}^2) \\
|
|
* &= x_N^2 - x_0^2 - N(\bar{x}_{i+1} - \bar{x}_{i})(\bar{x}_{i+1} + \bar{x}_{i})
|
|
* \end{align*}
|
|
* ```
|
|
*
|
|
* - Recognizing that the difference of means can be expressed
|
|
*
|
|
* ```tex
|
|
* \bar{x}_{i+1} - \bar{x}_i = \frac{1}{N} \biggl( \sum_{k=1}^N x_k - \sum_{k=0}^{N-1} x_k \biggr) = \frac{x_N - x_0}{N}
|
|
* ```
|
|
*
|
|
* and substituting into the equation above
|
|
*
|
|
* ```tex
|
|
* (N-1)(\Delta s^2) = x_N^2 - x_0^2 - (x_N - x_0)(\bar{x}_{i+1} + \bar{x}_{i})
|
|
* ```
|
|
*
|
|
* - Rearranging terms gives us the update equation
|
|
*
|
|
* ```tex
|
|
* \begin{align*}
|
|
* (N-1)(\Delta s^2) &= (x_N - x_0)(x_N + x_0) - (x_N - x_0)(\bar{x}_{i+1} + \bar{x}_{i})
|
|
* &= (x_N - x_0)(x_N + x_0 - \bar{x}_{i+1} - \bar{x}_{i}) \\
|
|
* &= (x_N - x_0)(x_N - \bar{x}_{i+1} + x_0 - \bar{x}_{i})
|
|
* \end{align*}
|
|
* ```
|
|
*
|
|
* @param {Collection} [out] - output array
|
|
* @param {PositiveInteger} window - window size
|
|
* @throws {TypeError} output argument must be array-like
|
|
* @throws {TypeError} window size must be a positive integer
|
|
* @returns {Function} accumulator function
|
|
*
|
|
* @example
|
|
* var accumulator = incrmmeanvar( 3 );
|
|
*
|
|
* var v = accumulator();
|
|
* // returns null
|
|
*
|
|
* v = accumulator( 2.0 );
|
|
* // returns [ 2.0, 0.0 ]
|
|
*
|
|
* v = accumulator( -5.0 );
|
|
* // returns [ -1.5, 24.5 ]
|
|
*
|
|
* v = accumulator( 3.0 );
|
|
* // returns [ 0.0, 19.0 ]
|
|
*
|
|
* v = accumulator( 5.0 );
|
|
* // returns [ 1.0, 28.0 ]
|
|
*
|
|
* v = accumulator();
|
|
* // returns [ 1.0, 28.0 ]
|
|
*/
|
|
function incrmmeanvar( out, window ) {
|
|
var meanvar;
|
|
var delta;
|
|
var buf;
|
|
var tmp;
|
|
var M2;
|
|
var mu;
|
|
var d1;
|
|
var d2;
|
|
var W;
|
|
var N;
|
|
var n;
|
|
var i;
|
|
if ( arguments.length === 1 ) {
|
|
meanvar = [ 0.0, 0.0 ];
|
|
W = out;
|
|
} else {
|
|
if ( !isArrayLike( out ) ) {
|
|
throw new TypeError( 'invalid argument. Output argument must be an array-like object. Value: `' + out + '`.' );
|
|
}
|
|
meanvar = out;
|
|
W = window;
|
|
}
|
|
if ( !isPositiveInteger( W ) ) {
|
|
throw new TypeError( 'invalid argument. Window size must be a positive integer. Value: `' + W + '`.' );
|
|
}
|
|
buf = new Float64Array( W );
|
|
n = W - 1;
|
|
M2 = 0.0;
|
|
mu = 0.0;
|
|
i = -1;
|
|
N = 0;
|
|
|
|
return accumulator;
|
|
|
|
/**
|
|
* If provided a value, the accumulator function returns updated accumulated values. If not provided a value, the accumulator function returns the current accumulated values.
|
|
*
|
|
* @private
|
|
* @param {number} [x] - input value
|
|
* @returns {(ArrayLikeObject|null)} output array or null
|
|
*/
|
|
function accumulator( x ) {
|
|
var k;
|
|
var v;
|
|
if ( arguments.length === 0 ) {
|
|
if ( N === 0 ) {
|
|
return null;
|
|
}
|
|
meanvar[ 0 ] = mu;
|
|
if ( N === 1 ) {
|
|
if ( isnan( M2 ) ) {
|
|
meanvar[ 1 ] = NaN;
|
|
} else {
|
|
meanvar[ 1 ] = 0.0;
|
|
}
|
|
} else if ( N < W ) {
|
|
meanvar[ 1 ] = M2 / (N-1);
|
|
} else {
|
|
meanvar[ 1 ] = M2 / n;
|
|
}
|
|
return meanvar;
|
|
}
|
|
// Update the index for managing the circular buffer:
|
|
i = (i+1) % W;
|
|
|
|
// Case: incoming value is NaN, the sliding second moment is automatically NaN...
|
|
if ( isnan( x ) ) {
|
|
N = W; // explicitly set to avoid `N < W` branch
|
|
mu = NaN;
|
|
M2 = NaN;
|
|
}
|
|
// Case: initial window...
|
|
else if ( N < W ) {
|
|
buf[ i ] = x; // update buffer
|
|
N += 1;
|
|
delta = x - mu;
|
|
mu += delta / N;
|
|
M2 += delta * (x - mu);
|
|
|
|
meanvar[ 0 ] = mu;
|
|
if ( N === 1 ) {
|
|
meanvar[ 1 ] = 0.0;
|
|
} else {
|
|
meanvar[ 1 ] = M2 / (N-1);
|
|
}
|
|
return meanvar;
|
|
}
|
|
// Case: N = W = 1
|
|
else if ( N === 1 ) {
|
|
mu = x;
|
|
M2 = 0.0;
|
|
meanvar[ 0 ] = x;
|
|
meanvar[ 1 ] = 0.0;
|
|
return meanvar;
|
|
}
|
|
// Case: outgoing value is NaN, and, thus, we need to compute the accumulated values...
|
|
else if ( isnan( buf[ i ] ) ) {
|
|
N = 1;
|
|
mu = x;
|
|
M2 = 0.0;
|
|
for ( k = 0; k < W; k++ ) {
|
|
if ( k !== i ) {
|
|
v = buf[ k ];
|
|
if ( isnan( v ) ) {
|
|
N = W; // explicitly set to avoid `N < W` branch
|
|
mu = NaN;
|
|
M2 = NaN;
|
|
break; // second moment is automatically NaN, so no need to continue
|
|
}
|
|
N += 1;
|
|
delta = v - mu;
|
|
mu += delta / N;
|
|
M2 += delta * (v - mu);
|
|
}
|
|
}
|
|
}
|
|
// Case: neither the current second moment nor the incoming value are NaN, so we need to update the accumulated values...
|
|
else if ( isnan( M2 ) === false ) {
|
|
tmp = buf[ i ];
|
|
delta = x - tmp;
|
|
d1 = tmp - mu;
|
|
mu += delta / W;
|
|
d2 = x - mu;
|
|
M2 += delta * (d1 + d2);
|
|
}
|
|
// Case: the current second moment is NaN, so nothing to do until the buffer no longer contains NaN values...
|
|
buf[ i ] = x;
|
|
|
|
meanvar[ 0 ] = mu;
|
|
meanvar[ 1 ] = M2 / n;
|
|
return meanvar;
|
|
}
|
|
}
|
|
|
|
|
|
// EXPORTS //
|
|
|
|
module.exports = incrmmeanvar;
|