GY-63_MS5611/libraries/Kurtosis

Kurtosis

Arduino library for calculating skewness and kurtosis of a dataset (statistics).

Description

Experimental

This library was written after a request how to calculate the kurtosis of a histogram. Diving into this question I learned Kurtosis is a statistical metric, somewhat related to the skewness ("is balanced") of that data set. The goal or kurtosis is to detect outliers (as far as I understand) in the data set.

Not a statistician by profession, I searched the internet and found several sources that described the kurtosis measurement. There appear to be multiple methods to determine the kurtosis of a data set, so there appears no agreed standard (see Wikipedia).

The code of this library is based upon the code of John D. Cook in the blog named "Computing skewness and kurtosis in one pass". Thanks for his kind permission! See - https://www.johndcook.com/blog/skewness_kurtosis/

I adapted the code on a few places to improve performance a bit. Furthermore I named the library Kurtosis as calculating that metric is the prime purpose of this library.

For a deep technical reader about kurtosis and skewness I refer to the Wikipedia.

Links to other information are welcome.

The code is not well tested or verified so use with care.

Interpretation of Kurtosis metric.

• https://en.wiktionary.org/wiki/mesokurtosis et al.

First order interpretation.

Define: Excess Kurtosis is calculated as kurtosis - 3.

There are three levels of kurtosis.

• Mesokurtosis: An excess kurtosis of approx 0. The normal (Gaussian) distribution is mesokurtic
• Platykurtosis: A negative excess kurtosis. Platykurtic distributions have a "thin tail", they have relative few outliers.
• Leptokurtosis: A positive excess kurtosis. Leptokurtic distributions have a "thick tail", they have many outliers.

Interpretation of Skewness metric.

First order interpretation.

• Skewness between -0.5 and 0.5 => normal distribution
• Skewness between -1.0 and 1.0 => reasonable normal distribution
• Skewness smaller than -1.0 => left skewed distribution
• Skewness larger than 1.0 => right skewed distribution

Related

• https://github.com/RobTillaart/Histogram
• https://github.com/RobTillaart/Kurtosis
• https://github.com/RobTillaart/RunningAverage
• https://github.com/RobTillaart/RunningMedian
• https://github.com/RobTillaart/Statistic
• https://en.wikipedia.org/wiki/Skewness
• https://en.wikipedia.org/wiki/Kurtosis

Performance

Time in us, platform UNO,

0.0.0 is non optimized version.

function	0.0.0	0.1.0	0.1.1	Notes
add	232	213.7	211.6	averaged 1000 calls
variance	-	44	48	accuracy of UNO is 4 us.
skewness	416	352	356
kurtosis	-	64	64

Interface

#include "Kurtosis.h"

Base

• Kurtosis() create the Kurtosis object.
• void reset() resets the internal variables.

Core

• void add(double x) add new value to the internal variables.
• uint32_t count() returns the amount of values added.
• double mean() returns the mean (or average) of the values added. This function will return 0 if no values have been added, or after reset().
• double variance() idem. Note this value is cached as stddev() uses it too. So the performance of two consecutive calls differ in duration.
• double stddev() standard deviation.
• double skewness() returns skewness ("balanced"), negative values indicate left skew, positive values indicate right skew. See Wikipedia for details.
• double kurtosis() returns the kurtosis, "tailedness" is used e.g. to detect outliers. See Wikipedia for details.

Operators

For adding distributions.

• operator + add two Kurtosis objects together.
• operator += add a Kurtosis objects to this one.

debug / development

• void dump() dumps the internal variables to Serial.

Future

Must

• improve documentation.
• understand the metrics and math in far more detail.

Should

• add examples to explain the purpose.
• extend unit tests.
• ask the experts
• investigate the order of adding values on the metrics.

Could

• cache skewness and kurtosis? useful?
• Template class so one can choose float or double.

Wont

Support

If you appreciate my libraries, you can support the development and maintenance. Improve the quality of the libraries by providing issues and Pull Requests, or donate through PayPal or GitHub sponsors.

Thank you,