source: branches/0.4-stable/yat/statistics/Fisher.cc @ 1621

// $Id: Fisher.cc 1621 2008-11-10 21:39:48Z peter $

/*
  Copyright (C) 2004 Jari Häkkinen, Peter Johansson
  Copyright (C) 2005 Peter Johansson
  Copyright (C) 2006, 2007 Jari Häkkinen, Peter Johansson
  Copyright (C) 2008 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 2 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 this program; if not, write to the Free Software
  Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA
  02111-1307, USA.
*/

#include "Fisher.h"
#include "utility.h"

#include <gsl/gsl_cdf.h>
#include <gsl/gsl_randist.h>

namespace theplu {
namespace yat {
namespace statistics {

  
  Fisher::Fisher()
    : a_(0), b_(0), c_(0), d_(0), minimum_size_(10), oddsratio_(1.0)
  {
  }


  Fisher::~Fisher()
  {
  }


  bool Fisher::calculate_p_exact(void) const
  {
    return ( a_<minimum_size_ || b_<minimum_size_ || 
             c_<minimum_size_ || d_<minimum_size_); 
  }

  double Fisher::Chi2() const
  {
    double a,b,c,d;
    expected(a,b,c,d);
    return (a-a_)*(a-a_)/a + (b-b_)*(b-b_)/b + 
      (c-c_)*(c-c_)/c + (d-d_)*(d-d_)/d;
  }


  void Fisher::expected(double& a, double& b, double& c, double& d) const
  {
    double N = a_+b_+c_+d_;
    a =((a_+b_)*(a_+c_)) / N;
    b =((a_+b_)*(b_+d_)) / N;
    c =((c_+d_)*(a_+c_)) / N;
    d =((c_+d_)*(b_+d_)) / N;
  }


  unsigned int& Fisher::minimum_size(void)
  {
    return minimum_size_;
  }


  const unsigned int& Fisher::minimum_size(void) const
  {
    return minimum_size_;
  }


  double Fisher::oddsratio(const unsigned int a,
                           const unsigned int b,
                           const unsigned int c,
                           const unsigned int d) 
  {
    // If a column sum or a row sum is zero, the table is nonsense
    if ((a==0 || d==0) && (c==0 || b==0)){
      throw std::runtime_error("runtime_error: Table in Fisher is not valid\n");
    }
    a_ = a;
    b_ = b;
    c_ = c;
    d_ = d;

    oddsratio_=static_cast<double>(a*d)/static_cast<double>(b*c);
    return oddsratio_;
  }


  double Fisher::p_value() const
  {
    if (oddsratio_>1.0)
      return 2*p_value_one_sided();
    if (oddsratio_<1.0){
      // If oddsratio is less than unity
      // Two-sided p-value is
      // P(X <= oddsratio) + P(X >= 1/oddsratio) = 
      // P(X <= oddsratio) + P(X <= oddsratio)
      // 2 * (P(X < oddsratio) + P(X = oddsratio))
      // 2 * (1 - P(X >= oddsratio) + P(X = oddsratio))
      if (calculate_p_exact())
        return 2*(1-p_value_one_sided()+
                  gsl_ran_hypergeometric_pdf(a_,a_+b_,c_+d_, a_+c_)
                  );
      // if p is calculated approximatively correction is not needed
      return 2*(1-p_value_one_sided());

    }
    return 1.0;
  }


  double Fisher::p_value_one_sided() const
  {
    if ( calculate_p_exact() )
      return p_value_exact();
    return p_value_approximative();
  }


  double Fisher::p_value_approximative() const
  {
    return gsl_cdf_chisq_Q(Chi2(), 1.0);
  }

  double Fisher::p_value_exact() const
  { 
    // Since the calculation is symmetric and cdf_hypergeometric_P
    // loops up to k we choose the smallest number to be k and mirror
    // the matrix. This choice makes the p-value two-sided.

    if (a_<b_ && a_<c_ && a_<d_)
      return statistics::cdf_hypergeometric_P(a_,a_+b_,c_+d_,a_+c_);
    else if (b_<a_ && b_<c_ && b_<d_)
      return  statistics::cdf_hypergeometric_P(b_,a_+b_,c_+d_,b_+d_);
    else if (c_<a_ && c_<b_ && c_<d_)
      return  statistics::cdf_hypergeometric_P(c_,c_+d_,a_+b_,a_+c_);

    return statistics::cdf_hypergeometric_P(d_,c_+d_,a_+b_,b_+d_);

  }


}}} // of namespace statistics, yat, and theplu