sample distribution

takes a large and continuous data stream and continuously only retain a reduced empirical CDF approximation

small, simple, no dependencies

• Example • Features • Limitations • Why • API • License

Example

import Recorder from 'sample-distribution'

// recorder with 5 retained samples
var recorder = new Recorder(5)
[5,3,6,2,7,1,8].forEach(recorder.push, recorder)

console.log('minimum:', recorder.Q(0)) // minimum:0
console.log('median:', recorder.Q(0.5)) // median:4
console.log('maximum:', recorder.Q(1)) // maximum:8

Features

• constant memory use, no compression steps and/or triggered garbage collection
• significantly faster than other implementations (about 3-5x faster)
• mean-preserving compression (empirical cdf with same average as samples)
• No value interpolation. Values are kept as-is, but ranks are interpolated.

Limitations

• other than the mean, the other moments (variance, skew, kurtosis) are not preserved
• The moments are those of the generated curve that approximate those of the source samples

Background and related projects

• tdigest based on the work of Dunning
• h-digest based on the above but keeping values and adjusting ranks only

This module attempts to match the underlying CDF as closely as possible by adjusting local ranks to keep the local average when discarding values. By some measure, the root mean square error for each original sample value is 10 times better than the 2 other implementations above. Type npm run compare or see ./util/compare.js for a benchmark and error comparison for multiple distribution types.

API

Creation

• var recorder = new Recorder(L): creates a recorder that will keep L values and associated L ranks

Properties - read-only getters

• .N number: total samples received
• .E number: average of samples received and of the resulting approximated cdf
• .S number: standard variation of the resulting aproximated cdf
• .V number: variance of the resulting aproximated cdf

Properties - internal

• .vs array: internal store of retained sample values
• .rs array: internal store of retained value approximated ranks

Methods

• .push(number) void: sample value(s) to be added
• .F(value:number) number: cummulative probability (cdf) of the specified value
• .f(value:number) number: probability density (pdf) of the specified value
• .Q(probability:number) number: estimated value for specified probability
• .M(order:number) number: estimated origin moment (ie. E = M(1), V=M(2)-E^2)

Transferable through this.data

• const main = new Recorder( new Float64Array(buffer, offset, length) )
• const copy = new Recorder( main.data ) transferable TypedArray

License

MIT © Hugo Villeneuve