VarianceAggregator Class Reference

Author: Crownstone Team Copyright: Crownstone (https://crownstone.rocks) Date: Nov 29, 2020 License: LGPLv3+, Apache License 2.0, and/or MIT (triple-licensed) More...

#include <cs_Variance.h>

Public Member Functions
void	addValue (float new_measurement)
	update the aggregated data with a new measurement. More...
int	getCount () const
float	getMean () const
float	getVariance () const
float	getStandardDeviation () const
bool	isNumericPrecisionLow () const
void	reduceCount ()
	Reduce |M2| and num_recorded_values to prevent overflow. More...
void	reset ()

Private Attributes
uint32_t	num_recorded_values = 0
float	M2 = 0.0f
float	mean = 0.0f

Static Private Attributes
static const constexpr float	float_precision_threshold = 10e9f
static const constexpr int	count_precision_threshold = 10e6

Detailed Description

Author: Crownstone Team Copyright: Crownstone (https://crownstone.rocks) Date: Nov 29, 2020 License: LGPLv3+, Apache License 2.0, and/or MIT (triple-licensed)

Compute mean and variance of a running measurement without keeping track of all data points using this aggregation algorithm.

https://en.wikipedia.org/wiki/Algorithms_for_calculating_variance https://www.tandfonline.com/doi/abs/10.1080/00401706.1962.10490022

Notes:

• no optimization for catastrophic cancellation implemented
• no overflow protection implemented

Member Function Documentation

addValue()

void VarianceAggregator::addValue ( float  new_measurement )

inline

update the aggregated data with a new measurement.

getCount()

int VarianceAggregator::getCount ( ) const

inline

getMean()

float VarianceAggregator::getMean ( ) const

inline

getStandardDeviation()

float VarianceAggregator::getStandardDeviation ( ) const

inline

getVariance()

float VarianceAggregator::getVariance ( ) const

inline

isNumericPrecisionLow()

bool VarianceAggregator::isNumericPrecisionLow ( ) const

inline

reduceCount()

void VarianceAggregator::reduceCount ( )

inline

Reduce |M2| and num_recorded_values to prevent overflow.

The factors have been chosen to ensure that if we have: float v0 = getVariance(); reduceCount(); float v1 = getVariance(); Then the absolute difference |v0-v1| is minimal (0 if no rounding error occurs).

Since the mean converges to the actual mean of the measurements, no correction is necessary for that member value.

reset()

void VarianceAggregator::reset ( )

inline

Member Data Documentation

count_precision_threshold

const constexpr int VarianceAggregator::count_precision_threshold = 10e6

staticconstexprprivate

float_precision_threshold

const constexpr float VarianceAggregator::float_precision_threshold = 10e9f

staticconstexprprivate

M2

float VarianceAggregator::M2 = 0.0f

private

mean

float VarianceAggregator::mean = 0.0f

private

num_recorded_values

uint32_t VarianceAggregator::num_recorded_values = 0

private

---

The documentation for this class was generated from the following file:

• source/include/util/cs_Variance.h