source: trunk/yat/statistics/KolmogorovSmirnovOneSample.h @ 2998

#ifndef theplu_yat_statistics_kolmogorov_smirnov_one_sample
#define theplu_yat_statistics_kolmogorov_smirnov_one_sample

// $Id: KolmogorovSmirnovOneSample.h 2998 2013-03-14 08:08:54Z peter $

/*
  Copyright (C) 2013 Peter Johansson

  This file is part of the yat library, http://dev.thep.lu.se/yat

  The yat library is free software; you can redistribute it and/or
  modify it under the terms of the GNU General Public License as
  published by the Free Software Foundation; either version 3 of the
  License, or (at your option) any later version.

  The yat library is distributed in the hope that it will be useful,
  but WITHOUT ANY WARRANTY; without even the implied warranty of
  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
  General Public License for more details.

  You should have received a copy of the GNU General Public License
  along with yat. If not, see <http://www.gnu.org/licenses/>.
*/

#include <iosfwd>

#include <map>

namespace theplu {
namespace yat {
namespace statistics {

  /**
     \brief Kolmogorov Smirnov Test for one class

     Class can be used to test if data follows standard uniform
     distribution. Other distributions are not supported directly, but
     it is possible to check against other distribution by tranforming
     the data. If, for example, X is standard exponentially
     distributed log(X) is standard uniform. Therefore, rather than
     testing if X is exponential it is equivalent to test if log(X) is
     standard uniform.

     \since New in yat 0.11
   */
  class KolmogorovSmirnovOneSample
  {
  public:
    /**
       \brief Constructor
     */
    KolmogorovSmirnovOneSample(void);

    /**
       \brief add a value

       \a value should be a value in range [0, 1].
     */
    void add(double value, double weight=1.0);

    /**
       \brief Large-Sample Approximation

       This analytical approximation of p-value can be used when all
       weight equal unity and sample size \a n is large. The p-value
       is calcuated as \f$ P = \displaystyle - 2 \sum_{k=1}^{\infty}
       (-1)^ke^{-2k^2s^2}\f$, where s is the scaled score:

       \f$ s = \sqrt n \f$ score().
    */
    double p_value(void) const;

    /**
       \brief Remove a data point

       \throw utility::runtime_error if no data point exist with \a
       value and \a weight.
     */
    void remove(double value, double weight=1.0);

    /**
       \brief resets everything to zero
    */
    void reset(void);

    /**
       \brief Kolmogorov Smirnov statistic

       \f$ sup_x | F(x) - x | \f$ where is the empirical cumulative
       distribution \f$ F(x) = \frac{\sum_{i:x_i\le x}w_i}{ \sum w_i}
       \f$
    */
    double score(void) const;

    /**
       Same as score() but keeping the sign, in other words,
       abs(signed_score())==score()

       A positive score implies that values \em on \em average are
       smaller 0.5.
     */
    double signed_score(void) const;

  private:
    mutable double score_;
    mutable bool cached_;
    std::multimap<double, double> data_;
    double sum_w_;

    friend std::ostream& operator<<(std::ostream&,
                                    const KolmogorovSmirnovOneSample&);

    // using compiler generated copy and assignment
    //KolmogorovSmirnovOneSample(const KolmogorovSmirnovOneSample&);
    //KolmogorovSmirnovOneSample& operator=(const KolmogorovSmirnovOneSample&);
  };

}}} // of namespace theplu yat statistics

#endif