Skip to content

Commit 1a27628

Browse files
kgrytekgryte
committed
Add strided interface to compute the standard deviation using a one-pass textbook algorithm
1 parent 2f4ab58 commit 1a27628

37 files changed

+3684
-0
lines changed

‎lib/node_modules/@stdlib/stats/base/dstdevtk/README.md

+236
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,236 @@
1+
<!--
2+
3+
@license Apache-2.0
4+
5+
Copyright (c) 2020 The Stdlib Authors.
6+
7+
Licensed under the Apache License, Version 2.0 (the "License");
8+
you may not use this file except in compliance with the License.
9+
You may obtain a copy of the License at
10+
11+
http://www.apache.org/licenses/LICENSE-2.0
12+
13+
Unless required by applicable law or agreed to in writing, software
14+
distributed under the License is distributed on an "AS IS" BASIS,
15+
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
16+
See the License for the specific language governing permissions and
17+
limitations under the License.
18+
19+
-->
20+
21+
# dstdevtk
22+
23+
> Calculate the [standard deviation][standard-deviation] of a double-precision floating-point strided array using a one-pass textbook algorithm.
24+
25+
<section class="intro>">
26+
27+
The population [standard deviation][standard-deviation] of a finite size population of size `N` is given by
28+
29+
<!-- <equation class="equation" label="eq:population_standard_deviation" align="center" raw="\sigma = \sqrt{\frac{1}{N} \sum_{i=0}^{N-1} (x_i - \mu)^2}" alt="Equation for the population standard deviation."> -->
30+
31+
<div class="equation" align="center" data-raw-text="\sigma = \sqrt{\frac{1}{N} \sum_{i=0}^{N-1} (x_i - \mu)^2}" data-equation="eq:population_standard_deviation">
32+
<img src="" alt="Equation for the population standard deviation.">
33+
<br>
34+
</div>
35+
36+
<!-- </equation> -->
37+
38+
where the population mean is given by
39+
40+
<!-- <equation class="equation" label="eq:population_mean" align="center" raw="\mu = \frac{1}{N} \sum_{i=0}^{N-1} x_i" alt="Equation for the population mean."> -->
41+
42+
<div class="equation" align="center" data-raw-text="\mu = \frac{1}{N} \sum_{i=0}^{N-1} x_i" data-equation="eq:population_mean">
43+
<img src="" alt="Equation for the population mean.">
44+
<br>
45+
</div>
46+
47+
<!-- </equation> -->
48+
49+
Often in the analysis of data, the true population [standard deviation][standard-deviation] is not known _a priori_ and must be estimated from a sample drawn from the population distribution. If one attempts to use the formula for the population [standard deviation][standard-deviation], the result is biased and yields an **uncorrected sample standard deviation**. To compute a **corrected sample standard deviation** for a sample of size `n`,
50+
51+
<!-- <equation class="equation" label="eq:corrected_sample_standard_deviation" align="center" raw="s = \sqrt{\frac{1}{n-1} \sum_{i=0}^{n-1} (x_i - \bar{x})^2}" alt="Equation for computing a corrected sample standard deviation."> -->
52+
53+
<div class="equation" align="center" data-raw-text="s = \sqrt{\frac{1}{n-1} \sum_{i=0}^{n-1} (x_i - \bar{x})^2}" data-equation="eq:corrected_sample_standard_deviation">
54+
<img src="" alt="Equation for computing a corrected sample standard deviation.">
55+
<br>
56+
</div>
57+
58+
<!-- </equation> -->
59+
60+
where the sample mean is given by
61+
62+
<!-- <equation class="equation" label="eq:sample_mean" align="center" raw="\bar{x} = \frac{1}{n} \sum_{i=0}^{n-1} x_i" alt="Equation for the sample mean."> -->
63+
64+
<div class="equation" align="center" data-raw-text="\bar{x} = \frac{1}{n} \sum_{i=0}^{n-1} x_i" data-equation="eq:sample_mean">
65+
<img src="" alt="Equation for the sample mean.">
66+
<br>
67+
</div>
68+
69+
<!-- </equation> -->
70+
71+
The use of the term `n-1` is commonly referred to as Bessel's correction. Note, however, that applying Bessel's correction can increase the mean squared error between the sample standard deviation and population standard deviation. Depending on the characteristics of the population distribution, other correction factors (e.g., `n-1.5`, `n+1`, etc) can yield better estimators.
72+
73+
</section>
74+
75+
<!-- /.intro -->
76+
77+
<section class="usage">
78+
79+
## Usage
80+
81+
```javascript
82+
var dstdevtk = require( '@stdlib/stats/base/dstdevtk' );
83+
```
84+
85+
#### dstdevtk( N, correction, x, stride )
86+
87+
Computes the [standard deviation][standard-deviation] of a double-precision floating-point strided array `x` using a one-pass textbook algorithm.
88+
89+
```javascript
90+
var Float64Array = require( '@stdlib/array/float64' );
91+
92+
var x = new Float64Array( [ 1.0, -2.0, 2.0 ] );
93+
var N = x.length;
94+
95+
var v = dstdevtk( N, 1, x, 1 );
96+
// returns ~2.0817
97+
```
98+
99+
The function has the following parameters:
100+
101+
- **N**: number of indexed elements.
102+
- **correction**: degrees of freedom adjustment. Setting this parameter to a value other than `0` has the effect of adjusting the divisor during the calculation of the [standard deviation][standard-deviation] according to `N-c` where `c` corresponds to the provided degrees of freedom adjustment. When computing the [standard deviation][standard-deviation] of a population, setting this parameter to `0` is the standard choice (i.e., the provided array contains data constituting an entire population). When computing the corrected sample [standard deviation][standard-deviation], setting this parameter to `1` is the standard choice (i.e., the provided array contains data sampled from a larger population; this is commonly referred to as Bessel's correction).
103+
- **x**: input [`Float64Array`][@stdlib/array/float64].
104+
- **stride**: index increment for `x`.
105+
106+
The `N` and `stride` parameters determine which elements in `x` are accessed at runtime. For example, to compute the [standard deviation][standard-deviation] of every other element in `x`,
107+
108+
```javascript
109+
var Float64Array = require( '@stdlib/array/float64' );
110+
var floor = require( '@stdlib/math/base/special/floor' );
111+
112+
var x = new Float64Array( [ 1.0, 2.0, 2.0, -7.0, -2.0, 3.0, 4.0, 2.0 ] );
113+
var N = floor( x.length / 2 );
114+
115+
var v = dstdevtk( N, 1, x, 2 );
116+
// returns 2.5
117+
```
118+
119+
Note that indexing is relative to the first index. To introduce an offset, use [`typed array`][mdn-typed-array] views.
120+
121+
<!-- eslint-disable stdlib/capitalized-comments -->
122+
123+
```javascript
124+
var Float64Array = require( '@stdlib/array/float64' );
125+
var floor = require( '@stdlib/math/base/special/floor' );
126+
127+
var x0 = new Float64Array( [ 2.0, 1.0, 2.0, -2.0, -2.0, 2.0, 3.0, 4.0 ] );
128+
var x1 = new Float64Array( x0.buffer, x0.BYTES_PER_ELEMENT*1 ); // start at 2nd element
129+
130+
var N = floor( x0.length / 2 );
131+
132+
var v = dstdevtk( N, 1, x1, 2 );
133+
// returns 2.5
134+
```
135+
136+
#### dstdevtk.ndarray( N, correction, x, stride, offset )
137+
138+
Computes the [standard deviation][standard-deviation] of a double-precision floating-point strided array using a one-pass textbook algorithm and alternative indexing semantics.
139+
140+
```javascript
141+
var Float64Array = require( '@stdlib/array/float64' );
142+
143+
var x = new Float64Array( [ 1.0, -2.0, 2.0 ] );
144+
var N = x.length;
145+
146+
var v = dstdevtk.ndarray( N, 1, x, 1, 0 );
147+
// returns ~2.0817
148+
```
149+
150+
The function has the following additional parameters:
151+
152+
- **offset**: starting index for `x`.
153+
154+
While [`typed array`][mdn-typed-array] views mandate a view offset based on the underlying `buffer`, the `offset` parameter supports indexing semantics based on a starting index. For example, to calculate the [standard deviation][standard-deviation] for every other value in `x` starting from the second value
155+
156+
```javascript
157+
var Float64Array = require( '@stdlib/array/float64' );
158+
var floor = require( '@stdlib/math/base/special/floor' );
159+
160+
var x = new Float64Array( [ 2.0, 1.0, 2.0, -2.0, -2.0, 2.0, 3.0, 4.0 ] );
161+
var N = floor( x.length / 2 );
162+
163+
var v = dstdevtk.ndarray( N, 1, x, 2, 1 );
164+
// returns 2.5
165+
```
166+
167+
</section>
168+
169+
<!-- /.usage -->
170+
171+
<section class="notes">
172+
173+
## Notes
174+
175+
- If `N <= 0`, both functions return `NaN`.
176+
- If `N - c` is less than or equal to `0` (where `c` corresponds to the provided degrees of freedom adjustment), both functions return `NaN`.
177+
- Some caution should be exercised when using the one-pass textbook algorithm. Literature overwhelmingly discourages the algorithm's use for two reasons: 1) the lack of safeguards against underflow and overflow and 2) the risk of catastrophic cancellation when subtracting the two sums if the sums are large and the variance small. These concerns have merit; however, the one-pass textbook algorithm should not be dismissed outright. For data distributions with a moderately large standard deviation to mean ratio (i.e., **coefficient of variation**), the one-pass textbook algorithm may be acceptable, especially when performance is paramount and some precision loss is acceptable (including a risk of returning a negative variance due to floating-point rounding errors!). In short, no single "best" algorithm for computing the variance exists. The "best" algorithm depends on the underlying data distribution, your performance requirements, and your minimum precision requirements. When evaluating which algorithm to use, consider the relative pros and cons, and choose the algorithm which best serves your needs.
178+
179+
</section>
180+
181+
<!-- /.notes -->
182+
183+
<section class="examples">
184+
185+
## Examples
186+
187+
<!-- eslint no-undef: "error" -->
188+
189+
```javascript
190+
var randu = require( '@stdlib/random/base/randu' );
191+
var round = require( '@stdlib/math/base/special/round' );
192+
var Float64Array = require( '@stdlib/array/float64' );
193+
var dstdevtk = require( '@stdlib/stats/base/dstdevtk' );
194+
195+
var x;
196+
var i;
197+
198+
x = new Float64Array( 10 );
199+
for ( i = 0; i < x.length; i++ ) {
200+
x[ i ] = round( (randu()*100.0) - 50.0 );
201+
}
202+
console.log( x );
203+
204+
var v = dstdevtk( x.length, 1, x, 1 );
205+
console.log( v );
206+
```
207+
208+
</section>
209+
210+
<!-- /.examples -->
211+
212+
* * *
213+
214+
<section class="references">
215+
216+
## References
217+
218+
- Ling, Robert F. 1974. "Comparison of Several Algorithms for Computing Sample Means and Variances." _Journal of the American Statistical Association_ 69 (348). American Statistical Association, Taylor & Francis, Ltd.: 859–66. doi:[10.2307/2286154][@ling:1974a].
219+
220+
</section>
221+
222+
<!-- /.references -->
223+
224+
<section class="links">
225+
226+
[standard-deviation]: https://en.wikipedia.org/wiki/Standard_deviation
227+
228+
[@stdlib/array/float64]: https://github.com/stdlib-js/stdlib
229+
230+
[mdn-typed-array]: https://developer.mozilla.org/en-US/docs/Web/JavaScript/Reference/Global_Objects/TypedArray
231+
232+
[@ling:1974a]: https://doi.org/10.2307/2286154
233+
234+
</section>
235+
236+
<!-- /.links -->

‎lib/node_modules/@stdlib/stats/base/dstdevtk/benchmark/benchmark.js

+96
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,96 @@
1+
/**
2+
* @license Apache-2.0
3+
*
4+
* Copyright (c) 2020 The Stdlib Authors.
5+
*
6+
* Licensed under the Apache License, Version 2.0 (the "License");
7+
* you may not use this file except in compliance with the License.
8+
* You may obtain a copy of the License at
9+
*
10+
* http://www.apache.org/licenses/LICENSE-2.0
11+
*
12+
* Unless required by applicable law or agreed to in writing, software
13+
* distributed under the License is distributed on an "AS IS" BASIS,
14+
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
15+
* See the License for the specific language governing permissions and
16+
* limitations under the License.
17+
*/
18+
19+
'use strict';
20+
21+
// MODULES //
22+
23+
var bench = require( '@stdlib/bench' );
24+
var randu = require( '@stdlib/random/base/randu' );
25+
var isnan = require( '@stdlib/math/base/assert/is-nan' );
26+
var pow = require( '@stdlib/math/base/special/pow' );
27+
var Float64Array = require( '@stdlib/array/float64' );
28+
var pkg = require( './../package.json' ).name;
29+
var dstdevtk = require( './../lib/dstdevtk.js' );
30+
31+
32+
// FUNCTIONS //
33+
34+
/**
35+
* Creates a benchmark function.
36+
*
37+
* @private
38+
* @param {PositiveInteger} len - array length
39+
* @returns {Function} benchmark function
40+
*/
41+
function createBenchmark( len ) {
42+
var x;
43+
var i;
44+
45+
x = new Float64Array( len );
46+
for ( i = 0; i < x.length; i++ ) {
47+
x[ i ] = ( randu()*20.0 ) - 10.0;
48+
}
49+
return benchmark;
50+
51+
function benchmark( b ) {
52+
var v;
53+
var i;
54+
55+
b.tic();
56+
for ( i = 0; i < b.iterations; i++ ) {
57+
v = dstdevtk( x.length, 1, x, 1 );
58+
if ( isnan( v ) ) {
59+
b.fail( 'should not return NaN' );
60+
}
61+
}
62+
b.toc();
63+
if ( isnan( v ) ) {
64+
b.fail( 'should not return NaN' );
65+
}
66+
b.pass( 'benchmark finished' );
67+
b.end();
68+
}
69+
}
70+
71+
72+
// MAIN //
73+
74+
/**
75+
* Main execution sequence.
76+
*
77+
* @private
78+
*/
79+
function main() {
80+
var len;
81+
var min;
82+
var max;
83+
var f;
84+
var i;
85+
86+
min = 1; // 10^min
87+
max = 6; // 10^max
88+
89+
for ( i = min; i <= max; i++ ) {
90+
len = pow( 10, i );
91+
f = createBenchmark( len );
92+
bench( pkg+':len='+len, f );
93+
}
94+
}
95+
96+
main();

0 commit comments

Comments
 (0)