Chi-Sqaured tests

This package helps run Chi-Squared hypothesis tests for testing distributions on numerical data. Details follow.

How to install

This library can easily be integrated in to your project manually. Alternatively, the library can be included using npm/GitHub packages.

npm install chi-sq-test

How to use

To run chi-squared test for a given dataset

ChiSqTest.gof(fObs, fExp, ddof)

Documentation

fObs: [Array] Array of observed frequencies for each category
Default: No default value, essential arg
fExp: [Array] Array of expected frequencies in each category
Default: By default all categories are assumed to be equally likely. Expected frequency of each category would be the mean of observed frequencies.
ddof: [number] delta degrees of freedom.
Effective degrees of freedom = k - 1 - ddof, where k is the number of observed frequencies.
Default ddof: 0

This is somewhat similar to SciPy.

Example:

const ChiSqTest = require('chi-sq-test');

const obs = [2, 3, 4]; // observed frequencies 
const exp = [3, 4, 5]; // expected frequencies    
const ddof = 0;       // delta degree of freedom (Degree Of Freedom  = 3-1 = 2)

const testres1 = ChiSqTest.gof(obs, exp, ddof);
console.log(testres1);

/*
=> { value: 0.7833333333333332, pValue: 0.6759293959125221 }
*/

const testres2 = ChiSqTest.gof(obs); // mean fObs is used as fExp by default
console.log(testres2);

/*
=> { value: 0.6666666666666666, pValue: 0.7165313783925148 }
*/

Output:

Function gof returns a JSON object, which contains Chi-Square value and the pValue for the given dataset.

Chi-square statistics for independence

ChiSqTest.independence(fObs, ddof)

Documentation

fObs: [2D Array] 2D-Array of observed frequencies of interestcting categories T_ij = (A_i ∩ B_j)
Default: No default value, essential arg
ddof: [number] delta degrees of freedom.
Effective degrees of freedom = (k - 1).(m - 1) - ddof, where k and m are number of categories in sets A and B respectively.
Default ddof: 0

Example

Statement

We have an email-dataset which is divided in two ways. \ A = {with image, without images} \ B = {Spam, No Spam}

fObs(i,j)	With Images	Without Images
Spam	160	240
No Spam	140	460

For the null hypothesis: \ H₀: Email spam and image attachment are independent. \ H_A: being spam and image attachment are dependent

const ChiSqTest = require('chi-sq-test');
const obs = [
    [160, 240],
    [140, 460]
];

console.log(
    ChiSqTest.independence(obs, 0)
);

/*
=> { value: 31.746031746031747, pValue: 1.7570790822318827e-8 }
*/

Output:

Function independence returns a JSON object, which contains Chi-Square value and the pValue for the hypothesis for indpendence.

neeraj3029/chi-sq-test