source: trunk/lib/statistics/AveragerPair.h @ 295

// $Id: AveragerPair.h 295 2005-04-29 09:15:58Z peter $

#ifndef _theplu_statistics_averagerpair_
#define _theplu_statistics__averagerpair_

#include <c++_tools/statistics/Averager.h>

#include <cmath>
#include <utility>


namespace theplu{

  class gslapi::vector;

namespace statistics{
  ///
  /// Class for taking care of mean and covariance of two variables.
  ///
  class AveragerPair
  {
  public:

    ///
    /// Default constructor
    ///
    AveragerPair(void);
    
    ///
    /// Constructor taking sum of \a x , \a xx , \a y , \a yy , xy and
    /// number of pair of values \a n
    ///
    AveragerPair(const double x, const double xx, const double y, 
                 const double yy, const double xy, const long);

    ///
    /// Copy constructor
    ///
    AveragerPair(const AveragerPair&);
    
    ///
    /// Adding \a n pairs of data points with value \a x and \a y.
    ///
    inline void add(const double x, const double y, const long n=1)
      { x_.add(x,n); y_.add(y,n), xy_ += n*x*y; }

    ///
    /// \f$\frac{\sum_i (x_i-m_x)(y_i-m_y)}{\sqrt{\sum_i
    /// (x_i-m_x)^2\sum_i (y_i-m_y)^2}}\f$ 
    ///
    /// @return Concordence correlation coefficient.
    ///
    inline double ccc(void) const 
      { return ( (x_.variance() || y_.variance() || x_.mean()-y_.mean() ) ? 
                 ((2*covariance()) / 
                  ((x_.variance()+y_.variance()) +
                   (x_.mean()-y_.mean())*(x_.mean()-y_.mean()))) : 0); }
  
    ///
    /// \f$\frac{\sum_i (x_i-m_x)(y_i-m_y)}{\sqrt{\sum_i
    /// (x_i-m_x)^2\sum_i (y_i-m_y)^2}}\f$ 
    ///
    /// @return Pearson correlation coefficient.
    ///
    inline double correlation(void) const 
      { return ((x_.std()>0 && y_.std()>0) ? 
                (covariance() / (x_.std()*y_.std()) ) : 0); }
  
    ///
    /// The covariance is calculated using the \f$ (n-1) \f$
    /// correction, which means it is the best unbiased estimator of
    /// the covariance \f$ \frac{1}{N-1}\sum_i (x_i-m_x)(y_i-m_y)\f$,
    /// where \f$m\f$ is the mean.
    ///
    /// @return The covariance.
    ///
    inline double covariance(void) const 
      { return (n()>1) ? (xy_ - x_.sum_x()*y_.mean()) / (n()-1): 0; }
  
    ///
    /// @return The mean of xy.
    ///
    inline double mean_xy(void) const { return xy_/n(); }

    ///
    /// @return Average squared deviation between x and y \f$
    /// \frac{1}{N} \sum (x-y)^2 \f$
    ///
    inline double msd() const {return x_.mean_sqr()+y_.mean_sqr()-2*mean_xy();}

    ///
    /// @return The number of pair of data points.
    ///
    inline long n(void) const { return x_.n(); }

    ///
    /// Resets everything to zero
    ///
    inline void reset(void) { x_.reset(); y_.reset(); xy_=0.0; }

    ///
    /// @return The sum of xy.
    ///
    inline double sum_xy(void) const { return xy_; }

    ///
    /// @return \f$ \sum_i (x_i-m_x)(y_i-m_y)\f$
    ///
    inline double sum_xy_centered(void) const {return xy_-x_.sum_x()*y_.mean();}

    ///
    /// @return A const refencer to the averager object for x.
    ///
    inline const Averager& x_averager(void) const { return x_; }

    ///
    /// @return A const reference to the averager object for y
    ///
    inline const Averager& y_averager(void) const { return y_; }

    ///
    /// The assigment operator
    ///
    inline const AveragerPair& operator=(const AveragerPair& a)
      { x_=a.x_; y_=a.y_; xy_=a.xy_; return *this; }

    ///
    /// Operator to add another Averager
    ///
    const AveragerPair& operator+=(const AveragerPair&);

  private:
    Averager x_;
    Averager y_;
    double  xy_;

  };

}} // of namespace statistics and namespace theplu

#endif