(full documentation: https://strike65.github.io/SwiftyStats/docs/)

SwiftyStats is a framework written entirely in Swift that makes heavy use of generic types. SwiftyStats contains frequently used statistical procedures.

It is a framework that is regularly developed and has been created out of passion rather than necessity.

New in version 1.1.7

some less typos
parameters for Poisson and Binomial distribution
refactoring

Important

Due to many changes and improvements, it is very likely that projects created with older versions (earlier than 1.1.0) will need to be modified after an update to version 1.1.0. This is especially true for the introduction of namespaces. As a result, the calling conventions for probability distributions have changed. Legacy functions such as pdfStudentTDist(_:_:_:) have been declared deprecated and will be removed in a later release. The new calling convention looks like SSProbDist.StudentT.pdf().

Swift 5

Starting with version 1.1.1, SwiftyStats is compatible with Swift 5.0. If you need Swift 4.*, please use version 1.1.0.

Overview

SwiftyStats is based on the class SSExamine. This class encapsulates the data to be processed and provides descriptive statistics. In principle, an SSExamine instance can contain data of any kind. The available statistical indicators depend of course on the type of data.

The following "namespaces"/classes are provided (among others):

SSExamine (class)
SSHypothesisTesting (enum)
SSProbDist (enum)
SSDataFrame instances encapsulate datasets. You can imagine the structure of an SSDataFrame object as a table: The columns of the table correspond to the individual dataset and the rows correspond to the data of the dataset.
SSCrossTab contains a cross table with the usual structure (like a n x m matrix) and provides the statistics needed for frequency comparisons (Chi-square, Phi, residuals etc.).

There are several extensions to standard Swift types (Array, Floating point types, String ).

Using Xcode

The attached Xcode project contains four targets:

SwiftyStats (for macOS)

SwiftyStatsMobile (for iOS)

SwiftyStatsTests (Testsuite)

SwiftStatsCLTest (a command line demo)

Each target must be built individually (i.e. no dependencies are defined)!

In addition a Playground is added to

test the framework and

do prototyping

The Swift Type Checker problem

Due to the extensive support of generic types, the type checker runs hot and takes a long time to compile. Therefore the code doesn't look "nice" in some places, because "complex" expressions (like (z1 + z1 - w) / (z1 * w)) had to be simplified.

How to Install

CocoaPods (recommended if your are on a Mac)

CocoaPods is the preferred way to add SwiftyStats to your project:

$> cd <YOUR_PROJECT_FOLDER>
$> vi Podfile

Your Podfile should looks like:

target 'YOURPROJECT' do
use_frameworks!
pod 'SwiftyStats'  
end

If you have to use a Swift version earlier than 4.2, replace the line pod 'SwiftyStats' by:

...
pod 'SwiftyStats', '1.1.1'
...

Save your changes, run:

$ pod install

and wait.

Swift Package Manager (recommended for Linux)

The Linux package is still under development and is considered to be unstable.

Edit your Package.swift file:

import PackageDescription
// for Swift 5
let version = "1.1.6"
// for earlier versions:
// let version = "1.1.1"
let package = Package(
name: "<YOUR_PACKAGE_NAME>",
dependencies: [
.package(url: "https://github.com/strike65/SwiftyStats", from: version)
]
)

Rebuild your project.
For more information about the Swift Package Manager click here

Build Swift Module from command line

Clone the repo to a directory of your choice and change to the directory.

Swift must be present on your system!

make or make debug builds the module in debug-configuration
make release builds the module in release-configuration
make test builds the module and performs some tests
make clean resets the build-process

Tests

The integrated test suite uses numerical data for comparison with the results calculated by SwiftyStats. The comparison data were generated using common computer algebra systems and numerical software.

How to use

import SwiftyStats

// example data
let data: Array<Double> = [3.14,1.21,5.6]
// because our data are double valued items, the parameter "characterSet" is ignored
let test = SSExamine<Double, Double>.init(withObject: data, levelOfMeasurement: .interval, characterSet: nil)
// prints out the arithmetic mean
print("\(test.arithmeticMean)")
// you can use the class to analyze strings too:
let testString = "This string must be analyzed!"
// in this case, only characters contained in CharacterSet.alphanumerics are added
let stringAnalyze = VTExamine<String>(withObject: data, levelOfMeasurement: .nominal, characterSet: CharacterSet.alphanumerics)
print("\(stringAnalyze.frequency("i")")
// print out the 95% quantile of the Student T distribution
do {
let q = try SSProbDist.StudentT.quantile(p: 0.95, degreesOfFreedom: 21)
print("\(q)")
}
catch {
print(error.localizedDescription)
}

Probability distributions in general are defined within relatively narrow conditions expressed in terms of certain parameters such as "degree of freedom", "shape" or "mean". For each distribution there are the following functions defined:

cdf: Cumulative Distribution Function
pdf: Probability Density Function
quantile: Inverse CDF
para: returns a SSProbDistParams struct (mean, variance, skewness, kurtosis)

Please always check if NaN or nil is returned.

Obtainable Statistics (more to come)

SSExamine

arithmeticMean
mode
commonest
scarcest
quantile(q:)
quartile
geometricMean
harmonicMean
contraHarmonicMean
poweredMean(order:)
trimmedMean(alpha:)
winsorizedMean(alpha:)
gastwirth
median
product
logProduct
maximum
minimum
range
quartileDeviation
relativeQuartileDistance
midRange
interquartileRange
interquantileRange(lowerQuantile:upperQuantile:)
variance(type:)
standardDeviation(type:)
standardError
entropy
relativeEntropy
herfindahlIndex
conc
gini
giniNorm
CR()
normalCI(alpha:populationSD:)
studentTCI(alpha:)
meanCI
cv
coefficientOfVariation
meanDifference
medianAbsoluteDeviation(center:scaleFactor:)
meanAbsoluteDeviation(center:)
meanRelativeDifference
semiVariance(type:)
moment(r:type:)
autocorrelation(n:)
kurtosisExcess
kurtosis
kurtosisType
skewness
skewnessType
hasOutliers(testType:)
outliers(alpha:max:testType:)
isGaussian
testForDistribution(targetDistribution:)
boxWhisker
elementsAsString(withDelimiter:asRow:encloseElementsBy:)
elementsAsArray(sortOrder:)
uniqueElements(sortOrder:)
frequencyTable(sortOrder:)
cumulativeFrequencyTable(format:)
eCDF(_:)
smallestFrequency
largestFrequency

SSCrossTab

rowCount
columnCount
rowSums
columnSums
rowSum(row:)
rowSum(rowName:)
columnSum(column:)
columnSum(columnName:)
total
rowTotal()
colummTotal()
largestRowTotal()
largestCellCount(atColumn:)
largestCellCount(atRow:)
relativeTotalFrequency(row:column:)
relativeTotalFrequency(rowName:columnName:)
relativeRowFrequency(row:column:)
relativeColumnFrequency(row:column:)
relativeRowMarginFrequency(row:)
relativeColumnMarginFrequency(column:)
expectedFrequency(row:column:)
expectedFrequency(rowName:columnName:)
residual(row:column:)
standardizedResidual(row:column:)
adjustedResidual(row:column:)
degreesOfFreedom
chiSquare
chiSquareLikelihoodRatio
chiSquareYates
covariance
pearsonR
chiSquareMH
phi
ccont
coefficientOfContingency
cramerV
lambda_C_R()
lambda_R_C()
r0
r1

Probability Functions (cdf, pdf, quantile, para - see above)

Beta
Binomial
Cauchy
Erlang
Exponential
Chi Square (central, non-central)
F-RATIO (central, non central)
Gamma
GAUSSIAN
Laplace
Log Normal
Logistic
Pareto
Binomial
Poisson
Rayleigh
STUDENT's T
NON-CENTRAL T-DISTRIBUTION
TRIANGULAR
TRIANGULAR with two params
UNIFORM
Wald / Inverse Normal
Weibull
CIRCULAR DISTRIBUTION

SSHypothesisTesting

#####Equality of means

twoSampleTTest(data1:data2:alpha:)
twoSampleTTest(sample1:sample2:alpha:)
oneSampleTTest(sample:mean:alpha:)
oneSampleTTEst(data:mean:alpha:)
matchedPairsTTest(set1:set2:alpha:)
matchedPairsTTest(data1:data2:alpha:)
oneWayANOVA(data:alpha:)
oneWayANOVA(data:alpha:)
oneWayANOVA(dataFrame:alpha:)
multipleMeansTest(dataFrame:alpha:)
multipleMeansTest(data:alpha:)
multipleMeansTest(data:alpha:)
tukeyKramerTest(dataFrame:alpha:)
tukeyKramerTest(data:alpha:)
scheffeTest(dataFrame:alpha:)
bonferroniTest(dataFrame:)

#####Autocorrelation

autocorrelationCoefficient(array:lag:)
autocorrelation(array:)

#####NPAR tests

ksGoFTest(array:targetDistribution:)
adNormalityTest(data:alpha:)
mannWhitneyUTest(set1:set2:)
wilcoxonMatchedPairs(set1:set2:)
signTest(set1:set2:)
binomialTest(numberOfSuccess:numberOfTrials:probability:alpha:alternative:)
binomialTest(data:characterSet:testProbability:successCodedAs:alpha:alternative:)
binomialTest(data:testProbability:successCodedAs:alpha:alternative:)
kolmogorovSmirnovTwoSampleTest(set1:set2:alpha:)
waldWolfowitzTwoSampleTest(set1:set2:)
kruskalWallisHTest(data:alpha:)

#####Outliers

grubbsTest(array:alpha:)
esdOutlierTest(array:alpha:maxOutliers:testType:)

#####Randomness

runsTest(array:alpha:useCuttingPoint:userDefinedCuttingPoint:alternative:)

#####Equality of variances

bartlettTest(array:alpha:)
leveneTest(array:testType:alpha:)
chiSquareVarianceTest(array:nominalVariance:alpha:)
fTestVarianceEquality(data1:data2:alpha:)

LICENSE

This framework is published under the GNU GPL 3