xmplr -- Exemplar -- Statistical object generator

This library creates "Exemplars" or statistical object generator for defining and creating an arbitrary object where each attribute conforms to a statistical distribution.

Note - This library creates random fictitious people

All of the people created are randomly generated and fictitious. Any resemblance to a real person is coincidental (obviously).

At some point, this project will add address, social security numbers, credit cards, etc to facilitate testing.

Example - generate a single customer at a time (typescript)

Generate customers as an xmplr Object model that will have a random but reasonable: first name, last name, adult age, date of purchase, and model from a list.

    let personModel = {
        first: new xFirstNames(),
        last: new xLastNames(),
        age: new xAdultAges(),
        location: new xZipCodes()
    }

    let xPerson = new xObject(personModel)
    let person = xPerson.next()
    console.log(person)

Generates a person

{ 
    first: 'Pamala',
    last: 'Berlin',
    age: 21,
    location: { 
      Zip: '78216', 
      Lat: '29.537264', 
      Lon: '-98.487882'
    } 
}

Example - an arrival rate of customers at a time

Return a list of models as "arrivals" satisfying a poisson distribution of number of arrivals per moment.

Wait for delayMsec milliseconds and return rate (10) per second. xRateRand/xRateModel work on millisecond rates, so divide you per second rate by 1000.

xRateModel takes a minimum arrival rate (1 in this case) to add to the randomly generated rate to act as a floor. This guarantees minimum number of arrivals per next() call.

async function arrivals(){
    let rate = 10/1000
    let xReceipts = new xRateModel(xPerson, new xRateRand(rate), 1)
    let delays = [100,400,1000,200]
    for( let i = 0; i < delays.length; i++ ){
        let delay = delays[i]
        await wait(delay)
        let receipts = xReceipts.next()
        assert( receipts.length > 0, `Receipts Has ${receipts.length} > 0` )
        let person:any = receipts[0]
        assert( typeof person.first === "string", `person has first name ${person.first}`)
    }
}
arrivals()

Simulating packet arrivals per second on a 1Gb network

1-Gb/s Ethernet interface can deliver between 80k and 1.4 million packets per second. An average network with larger packets may deliver about 100k packets per second. A xBetaRand shape (1,5) heavily weights in the first 40% and can approximate small office network traffic rates with spikes.

> import {xRateRand,xBetaRand} from "xmplr"
> var avg = 100000 // average of 100,000 packets per second
> var network = new xRateRand(avg,new x.xBetaRand(1,5))
> var pkts = () => Math.trunc(network.next())
> pkts()
22779
> pkts()
290768
> pkts()
7073
> pkts()
97729

Simulation Data

There are a number of US related source files in the "./data" directory.

Simulating names using Common US First and Last Names

The default names are contained in "First_Names.csv" and "last_Names.csv" files in the "./data" directory.

The names were downloaded from here

No license information was provided.

Supply your own array to xList() to create your own list of names, with possibly distributions.

Simulating locations using US Zip Codes

The default zip codes are contained in "US_Zip_Codes.csv" in the "./data" directory.

The data was downloaded from github user erichurst

Original source information:

All US zip codes with their corresponding latitude and longitude coordinates. Comma delimited for your database goodness.

Census Source

Monty Hall Problem

An example of the Monty Hall Problem is test/montyhall.tape.ts.

The test demonstrates a simple usage of xLists to simulate the Monty Hall Problem.

1. Behind 3 doors, place 1 car and 2 goats
2. Player chooses a door (in this game we "don't look" at our choice until the end)
3. Game show host shows a goat behind one of the remaining doors.
4. You can Stay on your first choice, or Switch to the other door.

Switching improves your odds from 1/3 to 2/3!

Run this test and see for yourself.