average.h

Go to the documentation of this file.

/*
 * Copyright (c) 2009 IITP RAS
 *
 * SPDX-License-Identifier: GPL-2.0-only
 *
 * Authors: Pavel Boyko <boyko@iitp.ru>
 * Corrections and extensions: Timo Bingmann <tbns@idlebox.net>
 */
 
#ifndef AVERAGE_H
#define AVERAGE_H
 
#include "basic-data-calculators.h"
 
#include <cmath>
#include <limits>
#include <ostream>
#include <stdint.h>
 
namespace ns3
{
 
/**
 * \ingroup stats
 *
 * Simple average, min, max and std. deviation calculator
 *
 */
 
template <typename T = double>
class Average
{
  public:
    Average()
        : m_size(0),
          m_min(std::numeric_limits<T>::max()),
          m_max(0)
    {
    }
 
    /**
     * Add new sample
     * \param x The sample
     */
    void Update(const T& x)
    {
        // Give the variance calculator the next value.
        m_varianceCalculator.Update(x);
 
        m_min = std::min(x, m_min);
        m_max = std::max(x, m_max);
        m_size++;
    }
 
    /// Reset statistics
    void Reset()
    {
        m_varianceCalculator.Reset();
 
        m_size = 0;
        m_min = std::numeric_limits<T>::max();
        m_max = 0;
    }
 
    // Sample statistics
    /**
     * Sample size
     * \return the sample size
     */
    uint32_t Count() const
    {
        return m_size;
    }
 
    /**
     * Sample minimum
     * \return the minimum of the sample
     */
    T Min() const
    {
        return m_min;
    }
 
    /**
     * Sample maximum
     * \return the maximum of the sample
     */
    T Max() const
    {
        return m_max;
    }
 
    /**
     * Sample average
     * \return the average of the sample
     */
    double Avg() const
    {
        return m_varianceCalculator.getMean();
    }
 
    /**
     * Sample estimate of mean, alias to Avg
     * \return the average of the sample
     */
    double Mean() const
    {
        return Avg();
    }
 
    /**
     * Sample unbiased nbiased estimate of variance
     * \return the unbiased nbiased estimate of variance
     */
    double Var() const
    {
        return m_varianceCalculator.getVariance();
    }
 
    /**
     * Sample standard deviation
     * \return the standard deviation
     */
    double Stddev() const
    {
        return std::sqrt(Var());
    }
 
    /**
     * \name Error of the mean estimates
     *
     * @{
     */
    /**
     * \brief Margin of error of the mean for 90% confidence level
     *
     * Note that estimates are valid for
     *   - uncorrelated measurements,
     *   - normal distribution and
     *   - large enough sample size.
     *
     * \returns Margin of error of the mean for 90% confidence level
     */
    double Error90() const
    {
        return 1.645 * std::sqrt(Var() / Count());
    }
 
    /**
     * \brief Margin of error of the mean for 95% confidence level
     *
     * Note that estimates are valid for
     *   - uncorrelated measurements,
     *   - normal distribution and
     *   - large enough sample size.
     *
     * \returns Margin of error of the mean for 95% confidence level
     */
    double Error95() const
    {
        return 1.960 * std::sqrt(Var() / Count());
    }
 
    /**
     * \brief Margin of error of the mean for 99% confidence level
     *
     * Note that estimates are valid for
     *   - uncorrelated measurements,
     *   - normal distribution and
     *   - large enough sample size.
     *
     * \returns Margin of error of the mean for 99% confidence level
     *
     */
    double Error99() const
    {
        return 2.576 * std::sqrt(Var() / Count());
    }
 
    /**@}*/
 
  private:
    uint32_t m_size;                                       //!< Number of sampled data.
    T m_min;                                               //!< Minimum value observed.
    T m_max;                                               //!< Maximum value observed.
    MinMaxAvgTotalCalculator<double> m_varianceCalculator; //!< Variance calculator.
};
 
/**
 * Print avg (err) [min, max]
 * \param os The output stream
 * \param x The Average value to print
 * \return the output stream.
 */
template <typename T>
std::ostream&
operator<<(std::ostream& os, const Average<T>& x)
{
    if (x.Count() != 0)
    {
        os << x.Avg() << " (" << x.Stddev() << ") [" << x.Min() << ", " << x.Max() << "]";
    }
    else
    {
        os << "NA"; // not available
    }
    return os;
}
} // namespace ns3
#endif /* AVERAGE_H */