simple-statistics / simple-statistics

simple statistics for node & browser javascript

Simple Statistics

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
• 📈 Benchmarks
• 😗 A list of other statistics libraries

Installation

• I'm using Node.js, Webpack, Browserify, Rollup, or another module bundler, and install packages from npm.
  • First, install the simple-statistics module, using npm install simple-statistics, then include the code with require or import:
  • I use the require function to use modules in my project. (most likely)
    • When you use require, you have the freedom to assign the module to any variable name you want, but you need to specify the module's name exactly: in this case, 'simple-statistics'. The require method returns an object with all of the module's methods attached to it.
      

      var ss = require('simple-statistics')

  • I use import to use modules in my project. I'm probably using Babel, @std/esm, Webpack, or Rollup.
    • Import all functions under the ss object:

      import * as ss from 'simple-statistics'

      Include a specific named export:

      import {min} from 'simple-statistics'

      Simple statistics has only named exports for ES6.
• I'm using Deno.
  • Simple Statistics is published for Deno: https://deno.land/x/simplestatistics
• I'm not using a module bundler. I'm writing a web page, and want to include simple-statistics using a script tag.
  • I want to support all browsers
    • When you use simple-statistics from a script tag, you don't get to choose the variable name it is assigned to: simple-statistics will always become available globally as the variable ss. You can reassign this variable to another name if you want to, but doing so is optional.

      <script src='https://unpkg.com/[email protected]/dist/simple-statistics.min.js'>
      </script>

  • I want to use ES6 modules in a browser and I'm willing to only support new browsers to do it
    • This module works great with the ?module query parameter of unpkg. If you specify type='module' in your script tag, you'll be able to import simple-statistics directly - through index.js and with true ES6 import syntax and behavior.

      <script type='module'>
      import {min} from "https://unpkg.com/[email protected]/index.js?module"
      console.log(min([1, 2, 3]))
      </script>

      This feature is still experimental in unpkg and very bleeding-edge.

simple-statistics's Issues

set up browser tests

Tag 1.0.0

This has been stable for a long time and I'm comfortable with the layout of the API. I should tag 1.0.0.

R-squared of a linear regression

Hello:

Mostly a question ... Is there an easy way to get the R-squared of a ss linear regression? I see that R-squared can be calculated separately, but wondering how to put the two together.

equal interval classification

Like jenks, but splits on equal intervals. Might be too trivial, but I have been finding myself building this out all the time when using ss.

Implement natural breaks

See https://github.com/simogeo/geostats/blob/master/lib/geostats.js#L407

k-means clustering

If stream-statistics has it, so should this

Bump version

Great library; thanks for the awesome work. If its not too much trouble, can you create a new tag (like 0.1.1 or whatever you think is appropriate). The main reason that I ask is that Bower doesn't really like handling branches that I can see. Many thanks.

equal number of features

Like jenks, but splits on equal number of features. I find this is often an alternative to jenks.

5 breaks:

[1,1,1,2,2,3,4,4,4,5]
=>
[1,1],[1,2],[2,3],[4,4],[4,5]

Add min/max and t_test to wiki

Streamy statistics

Many of the methods here would be amenable to stream-interface, so that they can adjust to new input. This is likely a new module.

Standard score

http://en.wikipedia.org/wiki/Standard_score

Implement linear transformations

http://en.wikipedia.org/wiki/Data_transformation_(statistics)

Jenks zero index?

Question - in the bits looping over the arrays I notice item 1 is set and all the loops go to len + 1.

Is this a side effect of fortran arrays starting at 1 ?

On a quick look, it looks like the zeroth elements are only accessed when the matrices are cleared.

http://developeronline.blogspot.co.uk/2008/04/why-array-index-should-start-from-0.html

This is probably just me misunderstanding the maths though :)

histogram buckets

https://github.com/substack/node-lump is pretty useful! i use it and cross filter here https://github.com/maxogden/datacouch/blob/master/attachments/script/application.js#L114

Comment quantile_sorted

Add new array methods to mixin constructor

skewness should be an array method as well!

Add CONTRIBUTING.md

This should make contribution comments easier.

Weighted linear regression

Hi

Do you, by any chance, know where I can find a Javascript library for weighted linear regression? If not, would you consider adding it to simple-statistics?

Thanks

Marcus

More tests and reference data for sample correlation

split into copy/pasteable functions

Document bayesian classifier

Document normal distribution

question : variance or sample variance?

I read your code and I didn't understand what is the difference between variance and sample variance. is it just the different value of denominator? I would expect that sample variance is a variance calculated on a subset.

Thanks

Add tests

Create homepage

This has nothing in index.js: it should have an intro or something.

Implement Skewness

in r and on wikipedia

Implement Harmonic mean

https://en.wikipedia.org/wiki/Harmonic_mean

expose a and b parameters for linear regression

Improve covariance documentation

We could do a better job of commenting and explaining covariance in the code.

Are sample variance and variance the same?

Looks like these are just the same algo...

Switch to tape

time to do it.

Linear regression example

Hi Tom,

The linear regression example at the bottom of the page doesn't seem to be working (http://bl.ocks.org/tmcw/3931800). Can you direct me to another solid simple example?

Thanks!
Kenan

RMS and general power means

any plans on supporting more general power means (such as root mean squared)?

Improve readme

Run jshint on travis

Easy enough, and makes contributions easier.

Implement lowess

http://en.wikipedia.org/wiki/Local_regression

Have a demo of integration with d3

Implement min/max

Obviously simple things, but they should be there.

Low-variance shifted data

Hi, please consider improving the variance calculation to handle low-variance floating point numbers:

https://en.wikipedia.org/wiki/Algorithms_for_calculating_variance#Computing_shifted_data

From #82

test cumulative prob

AMD

it's simple enough to support AMD too, should be a 3-line change.

Implement median absolute deviation (robust)

https://en.wikipedia.org/wiki/Median_absolute_deviation

Implement confidence bands

• http://en.wikipedia.org/wiki/Confidence_band
• http://en.wikipedia.org/wiki/Bonferroni_method

Quantile not quantiling

@tmcw @morganherlocker As is my wont, I have probably screwed up something basic, but I've been using ss_quantile() in what I think is the correct manner to classify data on a map, and I'm not getting equal numbers per bin.

Here's the implementation (adapted from the now-classic Leaflet choropleth example of yore):

// FORMAT COLOR SCHEME BY QUANTILE
function getColorPositive(d) {
  return d > 100 ? 'rgba(0,0,0,0)'
    : d >= ss.quantile(range,0.8) && d < 100 ? '#4682b4'
    : d >= ss.quantile(range,0.6) && d < ss.quantile(range,0.8) ? '#7a9fc7'
    : d >= ss.quantile(range,0.4) && d < ss.quantile(range,0.6) ? '#a7bfd9'
    : d >= ss.quantile(range,0.2) && d < ss.quantile(range,0.4) ? '#d4deec'
    : '#ffffff';
}

Here's an example of the range array:

[0.2271, 0.1937, 0.2017, 0.1092, 0.1401, 0.2652, 0.1757, undefined, 0.2026, 0.193, 0.1526, 0.3377, undefined, 0.2196, 0.1244, 0.1393, 0.1792, 0.4396, 0.1892, 0.2097, 0.2016, 0.1989, 0.259, undefined, undefined, 0.1836, undefined, 0.1684, 0.1412, 0.2843, undefined, undefined, 0.182, 0.176, 0.1228, 0.1085, 0.1032, 0.299, 0.1762, 0.2347, 0.2213, 0.333, 0.294, 0.2503, 0.2965, 0.2017, 0.1781, undefined, 0.2886, 0.2557, 0.1859, 0.2629, 0.1728, 0.1331, 0.26, 0.1627]`

And here's what it looks like on the map (with the above range):

. . . there are three of the dark red-binned features and twelve of the white-binned features; there should be 8-9 features per color bin. This happens pretty consistently for all the red-colored variables in the app.

What did I screw up?

Demo page doesn't work in FireFox

http://macwright.org/simple-statistics/ works in Chrome beta, but not Fx 24 or 34 Win7-64.
TIA :)

docs for sample_covariance() and sample_correlation() wrong

These two functions both take two arguments, but are documented as only having one.

I'd submit a pull request to fix it but I'm not literate in the documentation generation system.

Document 2-sample t test in API.md

tail issue of #43

Calculations incorrect.

console.log(stat.variance([1,2.0,3,4,5,6,7]))
4
console.log(stat.standard_deviation([1,2,3.0,4,5,6,7]))
2

sample function

Add sample function to reduce the time it takes to operate on larger data sets

split into files

there are enough functions in here that it makes sense to do this now.