source: trunk/lib/statistics/ROC.h @ 509

Last change on this file since 509 was 509, checked in by Peter, 16 years ago

added test for target
redesign crossSplitter
added two class function in Target

  • Property svn:eol-style set to native
  • Property svn:keywords set to Author Date Id Revision
File size: 3.4 KB
Line 
1// $Id: ROC.h 509 2006-02-18 13:47:32Z peter $
2
3#ifndef _theplu_statistics_roc_
4#define _theplu_statistics_roc_
5
6#include <c++_tools/classifier/Target.h>
7#include <c++_tools/statistics/Score.h>
8
9#include <utility>
10#include <vector>
11
12namespace theplu {
13namespace statistics { 
14
15  ///
16  /// Class for ROC (Reciever Operating Characteristic).
17  ///   
18  class ROC : public Score
19  {
20 
21  public:
22    ///
23    /// Default constructor
24    ///
25    ROC(bool absolute=true);
26         
27    ///
28    /// Destructor
29    ///
30    virtual ~ROC(void) {};
31         
32    /// Function taking \a value, \a target (+1 or -1) and vector
33    /// defining what samples to use. The score is equivalent to
34    /// Mann-Whitney statistics.
35    /// @return the area under the ROC curve. If the area is less
36    /// than 0.5 and absolute=true, 1-area is returned. Complexity is
37    /// \f$ N\log N \f$ where \f$ N \f$ is number of samples.
38    ///
39    double score(const classifier::Target& target, 
40                 const gslapi::vector& value); 
41   
42    /// Function taking values, target, weight and a vector defining
43    /// what samples to use. The area is defines as \f$ \frac{\sum
44    /// w^+w^-}{\sum w^+w^-}\f$, where the sum in the numerator goes
45    /// over all pairs where value+ is larger than value-. The
46    /// denominator goes over all pairs. If target is equal to 1,
47    /// sample belonges to class + otherwise sample belongs to class
48    /// -. @return wheighted version of area under the ROC curve. If
49    /// the area is less than 0.5 and absolute=true, 1-area is
50    /// returned. Complexity is \f$ N^2 \f$ where \f$ N \f$ is number
51    /// of samples.
52    ///
53    double score(const classifier::Target& target, 
54                 const gslapi::vector& value, 
55                 const gslapi::vector& weight); 
56       
57
58    ///
59    ///Calculates the p-value, i.e. the probability of observing an
60    ///area equally or larger if the null hypothesis is true. If P is
61    ///near zero, this casts doubt on this hypothesis. The null
62    ///hypothesis is that the values from the 2 classes are generated
63    ///from 2 identical distributions. The alternative is that the
64    ///median of the first distribution is shifted from the median of
65    ///the second distribution by a non-zero amount. If the smallest
66    ///group size is larger than minimum_size (default = 10), then P
67    ///is calculated using a normal approximation.  @return the
68    ///one-sided p-value( if absolute true is used this is equivalent
69    ///to the two-sided p-value.)
70    ///
71    double p_value(void) const;
72   
73    ///
74    /// minimum_size is the threshold for when a normal
75    /// approximation is used for the p-value calculation.
76    ///
77    /// @return reference to minimum_size
78    ///
79    inline u_int& minimum_size(void){ return minimum_size_; } 
80
81    bool target(const size_t i) const;
82    inline size_t n(void) const { return vec_pair_.size(); }
83    inline size_t n_pos(void) const { return nof_pos_; }
84
85  private:
86   
87    /// Implemented as in MatLab 13.1
88    double get_p_approx(const double) const;
89
90    /// Implemented as in MatLab 13.1
91    double get_p_exact(const double, const double, const double) const;
92
93    double area_;
94    u_int minimum_size_;
95    u_int nof_pos_;
96    std::vector<std::pair<bool, double> > vec_pair_; // class-value-pair
97  };
98
99  ///
100  /// The output operator for the ROC class. The output is an Nx2
101  /// matrix, where the first column is the sensitivity and second
102  /// is the specificity.
103  ///
104  std::ostream& operator<< (std::ostream& s, const ROC&);
105
106
107}} // of namespace statistics and namespace theplu
108
109#endif
110
Note: See TracBrowser for help on using the repository browser.