Typescript Math Toolkit Data Frame Analysis

I am currently working on an extension to the Typescript Math Toolkit that will be for clients only. This extension will enable front- or back-end (via NodeJS) developers to move analysis performed in R into browser-based, client-facing applications with minimal coding effort.

This is an alpha release of some of the (scaled-down) low-level utilities involved in that project. I wanted to at least share some of the development with the open-source community. Even if you are not interested in statistics, data science, or R, you might find something of interest here.

Author: Jim Armstrong - The Algorithmist

@algorithmist

theAlgorithmist [at] gmail [dot] com

Typescript: 2.0.3

Version: 1.0

Installation

Installation involves all the usual suspects

• npm and gulp installed globally
• Clone the repository
• npm install
• get coffee (this is the most important step)

Building and running the tests

1. gulp compile

2. gulp test

The test suite is in Mocha/Chai and specs reside in the test folder.

Contents

The current distribution consists of seven classes,

• TSMT$Chi2
• TSMT$DataStats (frequent visitors to my repos have seen this one)
• TSMT$Spcfcn
• TSMT$DataFrame
• TSMT$FrameStats
• TSMT$TableAnalysis
• TSMT$CompareUtils (you may find some useful utils for specs in this class)

Here is the public API for each class

TSMT$Chi2

get nu(): number
set nu(value: number)
p(x2: number): number
cdf(x2: number): number
q(x2: number): number
invCDF(p: number): number

TSMT$CompareUtils

static compare(a: number, b: number, d: number): boolean
static containsApproximately(value: number, compare: Array<number>, epsilon: number = 0.01): boolean
static vectorCompare(vector1: Array<number>, vector2: Array<number>, epsilon: number = 0.01): boolean

TSMT$DataFrame

The TSMT$DataFrame class also exports a union type for data that may be store in columns of a data frame.

export type tableData = string | number | boolean;

Methods:

fromArray(data: Array<Array<any>>, dataTypes: Array<number>): void
get categories(): Array<string>
get dataTypes(): Array<number>
get table(): Array<Array<tableData>>
get size(): number
getColumn(category: string): Array<tableData>
removeColumn(category: string): void
split(row: number, data?: Array<Array<any>>): Object

TSMT$DataStats

isEven(n: number): boolean
get samples(): number
get min(): number
get max(): number
get fiveNumbers(): Array<number>
get fences(): Object
getQuantile(p: number): Array<number>
get mean(): number
get geometricMean(): number
get harmonicMean(): number
get std(): number
get cv(): number
get median(): number
get mode(): number
getConfidenceInterval(_t: number): Object
get skewness(): number
get kurtosis(): number
covariance(x: Array<number>, y: Array<number>): number
correlation(x: Array<number>, y: Array<number>): number

TSMT$FrameStats

This class exports two enums that are used to describe statistics with one or two data columns as input and return a single number. The TSMT$DataStats class is used as a utility, but is never directly called by data frame code.

export enum SingleStatEnum
{
  MIN,
  MAX,
  RANGE,
  MEAN,
  MEDIAN,
  MODE,
  HARMONIC_MEAN,
  GEOMETRIC_MEAN,
  STDDEV,
  SKEWNESS,
  KURTOSIS
}

export enum DoubleStatEnum
{
  CORRELATION,
  COVARIANCE
}

Methods:

getSummary(frame: TSMT$DataFrame, category: string): Array<number>
singleStat(frame: TSMT$DataFrame, category: string, type: number): number
doubleStat(frame: TSMT$DataFrame, category1: string, category2: string, type: number): number
getFences(frame: TSMT$DataFrame, category: string): Object
getQuantiles(frame: TSMT$DataFrame, category: string, q: number): Array<number>
zScore(frame: TSMT$DataFrame): Array<Array<number>>

TSMT$Spcfcn

It is highly unlikely that any user would directly work with this class; its methods are provided for completeness.

All I have to say is thank God for 'Numerical Recipes' and numerous other publications on the numerical analysis and approximation of special functions :)

static incompleteBeta(a: number, b: number, x: number): number
static betacf(a: number, b: number, x:number ): number
static betaiapprox(a: number, b: number, x: number): number
static gammaln(_x: number): number
static gammp(a: number, x: number): number
static gammaq(a: number, x: number)
static gammapapprox(a: number, x: number, psig: number): number
static gser(a: number, x: number): number
static gcf(a: number, x: number): number
static invgammp(p: number, a: number): number

TSMT$TableAnalysis

oneWayTable(frame: TSMT$DataFrame, category: string, _asPercentage: boolean=false): Object
crossTable(frame: TSMT$DataFrame, category1: string, category2: string, grouping: Array<string>, colNames: Array<string>): Object
crossTabulation(frame: TSMT$DataFrame): Object
normalize(frame: TSMT$DataFrame): Array<Array<number>>
toArray(obj: Object): Array<Object>

Usage

A 'data frame' in rough R terms is akin to a 2D table of variant data. For purposes of this distribution, the top row of the table is the (string) name of each column. Cell data may be of type string, number, or boolean, although numeric data is the most common (and interesting).

A data frame is typically loaded into an instance of TSMT$DataFrame using the fromArray() method. Various statistical operations and analysis may be performed on columns of the table or across the entire table.

For demonstration purposes, I took the 'used cars' dataset used in the book 'Machine Learning in R' by Lantz and loaded it into a data frame, then performed some of the analysis that was done in the book.

NOTE: Category names are case sensitive!

Refer to the specs in the test folder for examples of using each of the classes.

License

Apache 2.0

Free Software? Yeah, Homey plays that