simple-statistics

What is simple-statistics?

The simple-statistics npm package provides a comprehensive set of statistical tools and functions for performing various statistical analyses. It is designed to be easy to use and covers a wide range of statistical operations, from basic descriptive statistics to more complex regression analysis.

What are simple-statistics's main functionalities?

Descriptive Statistics

This feature allows you to calculate basic descriptive statistics such as mean, median, and variance. The code sample demonstrates how to compute these statistics for a given dataset.

const ss = require('simple-statistics');
const data = [1, 2, 3, 4, 5, 6, 7, 8, 9, 10];
const mean = ss.mean(data);
const median = ss.median(data);
const variance = ss.variance(data);
console.log(`Mean: ${mean}, Median: ${median}, Variance: ${variance}`);

Probability Distributions

This feature provides functions to work with various probability distributions, including normal and binomial distributions. The code sample shows how to generate a random normal value and a binomial distribution.

const ss = require('simple-statistics');
const normal = ss.randomNormal(0, 1);
const binomial = ss.binomialDistribution(10, 0.5);
console.log(`Random Normal Value: ${normal}, Binomial Distribution: ${binomial}`);

Regression Analysis

This feature allows you to perform regression analysis, including linear regression. The code sample demonstrates how to compute a linear regression line for a given dataset.

const ss = require('simple-statistics');
const data = [[0, 1], [1, 2], [2, 3], [3, 4], [4, 5]];
const regression = ss.linearRegression(data);
const regressionLine = ss.linearRegressionLine(regression);
console.log(`Regression Line: y = ${regression.m}x + ${regression.b}`);

Hypothesis Testing

This feature provides functions for performing hypothesis testing, such as t-tests. The code sample shows how to perform a two-sample t-test on two datasets.

const ss = require('simple-statistics');
const tTest = ss.tTestTwoSample([1, 2, 3, 4, 5], [6, 7, 8, 9, 10]);
console.log(`T-Test Result: ${tTest}`);

Other packages similar to simple-statistics

jstat

jstat is a JavaScript library for statistical operations. It offers a wide range of statistical functions similar to simple-statistics, including descriptive statistics, probability distributions, and hypothesis testing. However, jstat has a more extensive API and supports matrix operations, making it suitable for more complex statistical analyses.

mathjs

mathjs is a comprehensive mathematics library for JavaScript and Node.js. It includes a variety of mathematical functions, including statistical operations. While it is not solely focused on statistics like simple-statistics, it provides a broader range of mathematical tools, making it a versatile choice for projects that require both statistical and general mathematical computations.

ssjs

ssjs (Statistical Software in JavaScript) is another library that provides statistical functions for JavaScript. It offers similar functionalities to simple-statistics, such as descriptive statistics and probability distributions. However, ssjs is less popular and has a smaller community compared to simple-statistics, which may affect the availability of support and updates.

README

A JavaScript implementation of descriptive, regression, and inference statistics.

Implemented in literate JavaScript with no dependencies, designed to work in all modern browsers (including IE) as well as in node.js.

API Documentation

---

Basic contracts of functions:

• Functions do not modify their arguments e.g. change their order
• Invalid input, like empty lists to functions that need 1+ items to work, will cause functions to return null.

Basic Array Operations

.mixin(array)

Optionally mix in the following functions into the Array prototype. Otherwise you can use them off of the simple-statistics object itself.

If given a particular array instance as an argument, this adds the functions only to that array rather than the global Array.prototype. Without an argument, it runs on the global Array.prototype.

.mean(x)

Mean of a single-dimensional Array of numbers. Also available as .average(x)

.sum(x)

Sum of a single-dimensional Array of numbers.

.mode(x)

Returns the number that appears most frequently in a single-dimensional Array of numbers. If there are multiple modes, the one that appears last is returned.

.variance(x)

Variance of a single-dimensional Array of numbers.

.standard_deviation(x)

Standard Deviation of a single-dimensional Array of numbers.

.median_absolute_deviation(x)

The Median Absolute Deviation (MAD) is a robust measure of statistical dispersion. It is more resilient to outliers than the standard deviation. Accepts a single-dimensional array of numbers and returns a dispersion value.

Also aliased to .mad(x) for brevity.

.median(x)

Median of a single-dimensional array of numbers.

.geometric_mean(x)

Geometric mean of a single-dimensional array of positive numbers.

.harmonic_mean(x)

Harmonic mean of a single-dimensional array of positive numbers.

.root_mean_square(x)

Root mean square (RMS) of a single-dimensional array of numbers.

.min(x)

Finds the minimum of a single-dimensional array of numbers. This runs in linear O(n) time.

.max(x)

Finds the maximum of a single-dimensional array of numbers. This runs in linear O(n) time.

.t_test(sample, x)

Does a student's t-test of a dataset sample, represented by a single-dimensional array of numbers. x is the known value, and the result is a measure of statistical significance.

.t_test_two_sample(sample_x, sample_y, difference)

The two-sample t-test is used to compare samples from two populations or groups, confirming or denying the suspicion (null hypothesis) that the populations are the same. It returns a t-value that you can then look up to give certain judgements of confidence based on a t distribution table.

This implementation expects the samples sample_x and sample_y to be given as one-dimensional arrays of more than one number each.

.sample_variance(x)

Produces sample variance of a single-dimensional array of numbers.

.sample_covariance(a, b)

Produces sample covariance of two single-dimensional arrays of numbers.

.sample_correlation(a, b)

Produces sample correlation of two single-dimensional arrays of numbers.

.quantile(sample, p)

Does a quantile of a dataset sample, at p. For those familiary with the k/q syntax, p == k/q. sample must be a single-dimensional array of numbers. p must be a number greater than or equal to zero and less than or equal to one, or an array of numbers following that rule. If an array is given, an array of results will be returned instead of a single number.

.chunk(sample, chunkSize)

Given a sample array, and a positive integer chunkSize, splits an array into chunks of chunkSize size and returns an array of those chunks. This does not change the input value. If the length of sample is not divisible by chunkSize, the last array will be shorter than the rest.

.quantile_sorted(sample, p)

Does a quantile of a dataset sample, at p. sample must be a one-dimensional sorted array of numbers, and p must be a single number greater than or equal to zero and less than or equal to one.

.iqr(sample)

Calculates the Interquartile range of a sample - the difference between the upper and lower quartiles. Useful as a measure of dispersion.

Also available as .interquartile_range(x)

.sample_skewness(sample)

Calculates the skewness of a sample, a measure of the extent to which a probability distribution of a real-valued random variable "leans" to one side of the mean. The skewness value can be positive or negative, or even undefined.

This implementation uses the Fisher-Pearson standardized moment coefficient, which means that it behaves the same as Excel, Minitab, SAS, and SPSS.

Skewness is only valid for samples of over three values.

.jenks(data, number_of_classes)

Find the Jenks Natural Breaks for a single-dimensional array of numbers as input and a desired number_of_classes. The result is a single-dimensional with class breaks, including the minimum and maximum of the input array.

.r_squared(data, function)

Find the r-squared value of a particular dataset, expressed as a two-dimensional Array of numbers, against a Function.

var r_squared = ss.r_squared([[1, 1]], function(x) { return x * 2; });

.cumulative_std_normal_probability(z)

Look up the given z value in a standard normal table to calculate the probability of a random variable appearing with a given value.

.z_score(x, mean, standard_deviation)

The standard score is the number of standard deviations an observation or datum is above or below the mean.

.standard_normal_table

A standard normal table from which to pull values of Φ (phi).

.error_function(x)

Calculate a numerical approximation to the Gaussian error function instead of looking up values in a table.

Regression

.linear_regression()

Create a new linear regression solver.

.data([[1, 1], [2, 2]])

Set the data of a linear regression. The input is a two-dimensional array of numbers, which are treated as coordinates, like [[x, y], [x1, y1]].

.line()

Get the linear regression line: this returns a function that you can give x values and it will return y values. Internally, this uses the m() and b() values and the classic y = mx + b equation.

var linear_regression_line = ss.linear_regression()
    .data([[0, 1], [2, 2], [3, 3]]).line();
linear_regression_line(5);

.m()

Just get the slope of the fitted regression line, the m component of the full line equation. Returns a number.

.b()

Just get the y-intercept of the fitted regression line, the b component of the line equation. Returns a number.

Classification

.bayesian()

Create a naïve bayesian classifier.

.train(item, category)

Train the classifier to classify a certain item, given as an object with keys, to be in a certain category, given as a string.

.score(item)

Get the classifications of a certain item, given as an object of category -> score mappings.

var bayes = ss.bayesian();
bayes.train({ species: 'Cat' }, 'animal');
bayes.score({ species: 'Cat' });
// { animal: 1 }

---

Literate Source

Usage

To use it in browsers, grab simple_statistics.js. To use it in node, install it with npm or add it to your package.json.

npm install simple-statistics

To use it with component,

component install tmcw/simple-statistics

To use it with bower,

bower install simple-statistics

Basic Descriptive Statistics

// Require simple statistics
var ss = require('simple-statistics');

// The input is a simple array
var list = [1, 2, 3];

// Many different descriptive statistics are supported
var sum = ss.sum(list),
    mean = ss.mean(list),
    min = ss.min(list),
    geometric_mean = ss.geometric_mean(list),
    max = ss.max(list),
    quantile = ss.quantile(0.25);

Linear Regression

// For a linear regression, it's a two-dimensional array
var data = [ [1, 2], [2, 3] ];

// simple-statistics can produce a linear regression and return
// a friendly javascript function for the line.
var line = ss.linear_regression()
    .data(data)
    .line();

// get a point along the line function
line(0);

var line = ss.linear_regression()

// Get the r-squared value of the line estimation
ss.r_squared(data, line);

Bayesian Classifier

var bayes = ss.bayesian();
bayes.train({ species: 'Cat' }, 'animal');
bayes.score({ species: 'Cat' });
// { animal: 1 }

Mixin Style

This is optional and not used by default. You can opt-in to mixins with ss.mixin().

This mixes simple-statistics methods into the Array prototype - note that extending native objects is a tricky move.

This will only work if defineProperty is available, which means modern browsers and nodejs - on IE8 and below, calling ss.mixin() will throw an exception.

// mixin to Array class
ss.mixin();

// The input is a simple array
var list = [1, 2, 3];

// The same descriptive techniques as above, but in a simpler style
var sum = list.sum(),
    mean = list.mean(),
    min = list.min(),
    max = list.max(),
    quantile = list.quantile(0.25);

Examples

• Linear regression with simple-statistics and d3js
• Jenks Natural Breaks with a choropleth map with d3js

Changelog

0.9.2

• Improved test coverage
• Switched linting from JSHint to eslint and fixed style issues this uncovered.