Online statistical algorithms for browsers and Node.js

Collection of online algorithms for data exploration and analysis implemented in JavaScript. Online algorithms process inputs piece by piece. That means you can process data without storing it in memory. More about online algorithms: https://en.wikipedia.org/wiki/Online_algorithm

Most algorithms of online-stats also support more usual batch mode (i.e. mean([1,2,3,4]))

Installation

npm i -S online-stats

Usage

const Stats = require('online-stats') // { Mean, Median, Max, Min, ... }

To process data sequentially we need functions to have internal state. That's why there's one extra step - functions initialization

const mean = Stats.Mean() // -> function mean
const median = Stats.Median() // -> function median
...

When functions are initialized, just call them passing a value (for example: mean(x)). Result is returned. To get a final result just call a function without any params: const result = mean()

Web demo

Check how the online-stats package is used for online data profiling on StatSim Profile

Mean

const mean = Stats.Mean()

mean(1) // -> 1
mean(2) // -> 1.5
mean(9) // -> 4

console.log(mean()) // -> 4

Variance

const v = Stats.Variance({ddof: 1}) // 0 (default) - population variance, 1 - sample variance

v(1) // -> 0
v(2) // -> 0.5
v(9) // -> 19

console.log(v()) // -> 19

Median

const median = Stats.Median()

median(1) // -> 1
median(2) // -> 1.5
median(9) // -> 2

console.log(median()) // -> 2

Min

const min = Stats.Min()

min(2) // -> 2
min(6) // -> 2
min(1) // -> 1

console.log(min()) // -> 1

Max

const max = Stats.Max()

max(2) // -> 2
max(6) // -> 6
max(1) // -> 6

console.log(max()) // -> 6

Standard Deviation

const std = Stats.Std({ddof: 1}) // 0 (default) - population std, 1 - sample std (unbiased)

std(1) // -> 0
std(2) // ~> 0.7071
std(9) // ~> 4.3589

console.log(std()) // ~> 4.3589

Covariance

const a = [1, 3, 2, 5, 8, 7, 12, 2, 4]
const b = [8, 6, 9, 4, 3, 3, 2, 7, 7]
const cov = Stats.Covariance({ddof: 1})

a.forEach((ax, i) => { cov(ax, b[i]) })
console.log(cov()) // -> -8.069

Histogram

const hist = Stats.Histogram(20)

hist(2)
hist(6)
hist(1)

console.log(hist())

Autocovariance

const autocov = Stats.AutoCov(5)

;[1, 2, 3, 4, 5, 6, 7].forEach(v => { autocov(v) })

console.log(autocov())

Autocorrelation

const autocor = Stats.AutoCor(5)

;[1, 2, 3, 4, 5, 6, 7].forEach(v => { autocor(v) })

console.log(autocor())

Linear regression

const lr = Stats.LinReg()

const f = x => 0.5 * x + 2
;[1, 2, 3, 4, 5, 6].forEach(v => { lr(v, f(v)) })

console.log(lr([7])) // Predict