statsbase v1.0.3

statsbase

An average, variance, covariance and correlation calculator written in TypeScript

lazy-stats TypeScript version initially.

• Example • Features • Limitations • API • License

Example

import Stats from 'statsbase';

const stat = new Stats(3); // for 3 random variables

stat.push(2, 1, 0);
stat.push([1, 1, 1]);
stat.push(0, 1, 2);

const average0 = stat.ave(0);
const average1 = stat.ave(1);
const variance2 = stat.var(2);
const covariance12 = stat.cov(1, 2);
const correlation20 = stat.cor(2, 0);

Features

• very small code and footprint for large number of instances
• only stores the summary values (average and covariances)
• uses Welford-style online single pass variance and covariance algorithm
• less than 100 sloc, no dependencies

Limitations

• all variables must have the same number of samples, pushed at the same time
• no skew and kurtosis

API

Properties

• .N number: total samples received
• .data Float64Array: transferable memory copy = new Stats( main.data )

Methods

• .push(number0, number1, ...) => {number} sampleSize - add sample value(s) and returns the sample size
• .push([number0, number1, ...]) => {number} sampleSize - add array of sample value(s) and returns the sample size
• .ave(index) => {number} - average of a given dataset. index is optional, if not provided, returns an array of all averages
• .var(index) => {number} - variance of a given dataset. index is optional, if not provided, returns an array of all variances
• .dev(index) => {number} - standard deviation of a given dataset. index is optional, if not provided, returns an array of all standard deviations
• .cov(j, i) => {number} - covariance between two datasets
• .cor(j, i) => {number} - correlation between two datasets
• .slope(j, i) => {number} - slope for y=set[j] and x=set[i]
• .intercept(j, i) => {number} - intercept for y=set[j] and x=set[i]
• .reset() => {object} this - clears all sums and counts back to 0

Creds

• lazy-stats The original repo.

License

Released under the MIT License