added correlation function to statistics (#3015)

* added correlation function to statistics

* added implemenation for signature Matrix, Matrix and support for BigNumbers

* reverted changes to default for version numbers of devDepenencies

* reverted changes to default for version numbers of devDepenencies in package-lock.json

* change variable name from xArray, yArray to x and y

* added Matrix as param in index.d.ts

* corrected the file and function names for correlation function

* renamed createCorrelation to createCorr in factoriesNumber.js

* fixed failing test case for matrix and added params and return in corr

AUTHORS

@@ -231,5 +231,6 @@ MaybePixem <47889538+MaybePixem@users.noreply.github.com>
 Aly Khaled <alykhaled2001@live.com>
 BuildTools <anikpatel1322@gmail.com>
 Anik Patel <74193405+Bobingstern@users.noreply.github.com>
+vrushaket <vrushu00@gmail.com>
 
 # Generated by tools/update-authors.js

src/expression/embeddedDocs/embeddedDocs.js

@@ -197,6 +197,7 @@ import { stdDocs } from './function/statistics/std.js'
 import { cumSumDocs } from './function/statistics/cumsum.js'
 import { sumDocs } from './function/statistics/sum.js'
 import { varianceDocs } from './function/statistics/variance.js'
+import { corrDocs } from './function/statistics/corr.js'
 import { acosDocs } from './function/trigonometry/acos.js'
 import { acoshDocs } from './function/trigonometry/acosh.js'
 import { acotDocs } from './function/trigonometry/acot.js'
@@ -543,6 +544,7 @@ export const embeddedDocs = {
   std: stdDocs,
   sum: sumDocs,
   variance: varianceDocs,
+  corr: corrDocs,
 
   // functions - trigonometry
   acos: acosDocs,

src/expression/embeddedDocs/function/statistics/corr.js

@@ -0,0 +1,22 @@
+export const corrDocs = {
+  name: 'corr',
+  category: 'Statistics',
+  syntax: [
+    'corr(A,B)'
+  ],
+  description: 'Compute the correlation coefficient of a two list with values, For matrices, the matrix correlation coefficient is calculated.',
+  examples: [
+    'corr([2, 4, 6, 8],[1, 2, 3, 6])',
+    'corr(matrix([[1, 2.2, 3, 4.8, 5], [1, 2, 3, 4, 5]]), matrix([[4, 5.3, 6.6, 7, 8], [1, 2, 3, 4, 5]]))'
+  ],
+  seealso: [
+    'max',
+    'mean',
+    'min',
+    'median',
+    'min',
+    'prod',
+    'std',
+    'sum'
+  ]
+}

src/factoriesAny.js

@@ -237,6 +237,7 @@ export { createMad } from './function/statistics/mad.js'
 export { createVariance } from './function/statistics/variance.js'
 export { createQuantileSeq } from './function/statistics/quantileSeq.js'
 export { createStd } from './function/statistics/std.js'
+export { createCorr } from './function/statistics/corr.js'
 export { createCombinations } from './function/probability/combinations.js'
 export { createCombinationsWithRep } from './function/probability/combinationsWithRep.js'
 export { createGamma } from './function/probability/gamma.js'

src/factoriesNumber.js

@@ -263,6 +263,7 @@ export { createMad } from './function/statistics/mad.js'
 export { createVariance } from './function/statistics/variance.js'
 export { createQuantileSeq } from './function/statistics/quantileSeq.js'
 export { createStd } from './function/statistics/std.js'
+export { createCorr } from './function/statistics/corr.js'
 
 // string
 export { createFormat } from './function/string/format.js'

src/function/statistics/corr.js

@@ -0,0 +1,64 @@
+import { factory } from '../../utils/factory.js'
+const name = 'corr'
+const dependencies = ['typed', 'matrix', 'mean', 'sqrt', 'sum', 'add', 'subtract', 'multiply', 'pow', 'divide']
+
+export const createCorr = /* #__PURE__ */ factory(name, dependencies, ({ typed, matrix, sqrt, sum, add, subtract, multiply, pow, divide }) => {
+  /**
+   * Compute the correlation coefficient of a two list with values, For matrices, the matrix correlation coefficient is calculated.
+   *
+   * Syntax:
+   *
+   *     math.corr(A, B)
+   *
+   * Examples:
+   *
+   *     math.corr([1, 2, 3, 4, 5], [4, 5, 6, 7, 8])     // returns 1
+   *     math.corr([1, 2.2, 3, 4.8, 5], [4, 5.3, 6.6, 7, 8])     // returns 0.9569941688503644
+   *     math.corr(math.matrix([[1, 2.2, 3, 4.8, 5], [1, 2, 3, 4, 5]]), math.matrix([[4, 5.3, 6.6, 7, 8], [1, 2, 3, 4, 5]])) // returns DenseMatrix [0.9569941688503644, 1]
+   *
+   * See also:
+   *
+   *     median, mean, min, max, sum, prod, std, variance
+   *
+   * @param {Array | Matrix} A The first array or matrix to compute correlation coefficient
+   * @param {Array | Matrix} B The second array or matrix to compute correlation coefficient
+   * @return {*} The correlation coefficient
+   */
+  return typed(name, {
+    'Array, Array': function (A, B) {
+      return _corr(A, B)
+    },
+    'Matrix, Matrix': function (xMatrix, yMatrix) {
+      return matrix(_corr(xMatrix.toArray(), yMatrix.toArray()))
+    }
+  })
+  /**
+   * Calculate the correlation coefficient between two arrays or matrices.
+   * @param {Array | Matrix} A
+   * @param {Array | Matrix} B
+   * @return {*} correlation coefficient
+   * @private
+   */
+  function _corr (A, B) {
+    if (Array.isArray(A[0]) && Array.isArray(B[0])) {
+      const correlations = []
+      for (let i = 0; i < A.length; i++) {
+        correlations.push(correlation(A[i], B[i]))
+      }
+      return correlations
+    } else {
+      return correlation(A, B)
+    }
+  }
+  function correlation (A, B) {
+    const n = A.length
+    const sumX = sum(A)
+    const sumY = sum(B)
+    const sumXY = A.reduce((acc, x, index) => add(acc, multiply(x, B[index])), 0)
+    const sumXSquare = sum(A.map(x => pow(x, 2)))
+    const sumYSquare = sum(B.map(y => pow(y, 2)))
+    const numerator = subtract(multiply(n, sumXY), multiply(sumX, sumY))
+    const denominator = sqrt(multiply(subtract(multiply(n, sumXSquare), pow(sumX, 2)), subtract(multiply(n, sumYSquare), pow(sumY, 2))))
+    return divide(numerator, denominator)
+  }
+})

test/unit-tests/function/statistics/corr.test.js

@@ -0,0 +1,21 @@
+import assert from 'assert'
+import math from '../../../../src/defaultInstance.js'
+const corr = math.corr
+const BigNumber = math.BigNumber
+
+describe('correlation', function () {
+  it('should return the correlation coefficient from an array', function () {
+    assert.strictEqual(corr([new BigNumber(1), new BigNumber(2.2), new BigNumber(3), new BigNumber(4.8), new BigNumber(5)], [new BigNumber(4), new BigNumber(5.3), new BigNumber(6.6), new BigNumber(7), new BigNumber(8)]).toNumber(), 0.9569941688503653)
+    assert.strictEqual(corr([1, 2, 3, 4, 5], [4, 5, 6, 7, 8]), 1)
+    assert.strictEqual(corr([1, 2.2, 3, 4.8, 5], [4, 5.3, 6.6, 7, 8]), 0.9569941688503644)
+    assert.deepStrictEqual(corr(math.matrix([[1, 2.2, 3, 4.8, 5], [1, 2, 3, 4, 5]]), math.matrix([[4, 5.3, 6.6, 7, 8], [1, 2, 3, 4, 5]]))._data, [0.9569941688503644, 1])
+  })
+
+  it('should throw an error if called with invalid number of arguments', function () {
+    assert.throws(function () { corr() })
+  })
+
+  it('should throw an error if called with an empty array', function () {
+    assert.throws(function () { corr([]) })
+  })
+})

types/index.d.ts

@@ -2970,6 +2970,14 @@ declare namespace math {
       normalization: 'unbiased' | 'uncorrected' | 'biased'
     ): MathNumericType
 
+    /**
+     * Calculate the correlation coefficient between two matrix.
+     * @param {Array | Matrix} x The first array or matrix to compute correlation coefficient
+     * @param {Array | Matrix} y The second array or matrix to compute correlation coefficient
+     * @returns correlation coefficient
+     */
+    corr(x: MathCollection, y: MathCollection): MathType
+
     /*************************************************************************
      * String functions
      ************************************************************************/