2642 |
10 Nov 11 |
peter |
1 |
#ifndef _theplu_yat_statistics_spearman_ |
2642 |
10 Nov 11 |
peter |
2 |
#define _theplu_yat_statistics_spearman_ |
2635 |
09 Nov 11 |
peter |
3 |
|
2635 |
09 Nov 11 |
peter |
// $Id$ |
2635 |
09 Nov 11 |
peter |
5 |
|
2635 |
09 Nov 11 |
peter |
6 |
/* |
2635 |
09 Nov 11 |
peter |
Copyright (C) 2011 Peter Johansson |
2635 |
09 Nov 11 |
peter |
8 |
|
2635 |
09 Nov 11 |
peter |
This file is part of the yat library, http://dev.thep.lu.se/yat |
2635 |
09 Nov 11 |
peter |
10 |
|
2635 |
09 Nov 11 |
peter |
The yat library is free software; you can redistribute it and/or |
2635 |
09 Nov 11 |
peter |
modify it under the terms of the GNU General Public License as |
2635 |
09 Nov 11 |
peter |
published by the Free Software Foundation; either version 3 of the |
2635 |
09 Nov 11 |
peter |
License, or (at your option) any later version. |
2635 |
09 Nov 11 |
peter |
15 |
|
2635 |
09 Nov 11 |
peter |
The yat library is distributed in the hope that it will be useful, |
2635 |
09 Nov 11 |
peter |
but WITHOUT ANY WARRANTY; without even the implied warranty of |
2635 |
09 Nov 11 |
peter |
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
2635 |
09 Nov 11 |
peter |
General Public License for more details. |
2635 |
09 Nov 11 |
peter |
20 |
|
2635 |
09 Nov 11 |
peter |
You should have received a copy of the GNU General Public License |
2635 |
09 Nov 11 |
peter |
along with yat. If not, see <http://www.gnu.org/licenses/>. |
2635 |
09 Nov 11 |
peter |
23 |
*/ |
2635 |
09 Nov 11 |
peter |
24 |
|
2648 |
14 Nov 11 |
peter |
25 |
#include <cstddef> |
2648 |
14 Nov 11 |
peter |
26 |
|
2635 |
09 Nov 11 |
peter |
27 |
namespace theplu { |
2635 |
09 Nov 11 |
peter |
28 |
namespace yat { |
2642 |
10 Nov 11 |
peter |
29 |
namespace statistics { |
2635 |
09 Nov 11 |
peter |
30 |
|
2635 |
09 Nov 11 |
peter |
31 |
/** |
2654 |
15 Nov 11 |
peter |
\brief Spearman rank correlation coefficient |
2654 |
15 Nov 11 |
peter |
33 |
|
2654 |
15 Nov 11 |
peter |
Class for Spearman rank correlation coefficient which can be |
2654 |
15 Nov 11 |
peter |
understood as Pearson correlation of ranks. |
2654 |
15 Nov 11 |
peter |
36 |
|
2654 |
15 Nov 11 |
peter |
\since New in yat 0.9 |
2635 |
09 Nov 11 |
peter |
38 |
*/ |
2635 |
09 Nov 11 |
peter |
39 |
class Spearman |
2635 |
09 Nov 11 |
peter |
40 |
{ |
2635 |
09 Nov 11 |
peter |
41 |
public: |
2635 |
09 Nov 11 |
peter |
42 |
/** |
2638 |
10 Nov 11 |
peter |
\brief Constructor |
2638 |
10 Nov 11 |
peter |
44 |
*/ |
2638 |
10 Nov 11 |
peter |
45 |
Spearman(void); |
2638 |
10 Nov 11 |
peter |
46 |
|
2638 |
10 Nov 11 |
peter |
47 |
/** |
2638 |
10 Nov 11 |
peter |
\brief Destructor |
2638 |
10 Nov 11 |
peter |
49 |
*/ |
2638 |
10 Nov 11 |
peter |
50 |
virtual ~Spearman(void); |
2638 |
10 Nov 11 |
peter |
51 |
|
2638 |
10 Nov 11 |
peter |
52 |
/** |
2635 |
09 Nov 11 |
peter |
\brief add pair of data |
2635 |
09 Nov 11 |
peter |
54 |
*/ |
2640 |
10 Nov 11 |
peter |
55 |
void add(double x, double y); |
2637 |
10 Nov 11 |
peter |
56 |
|
2635 |
09 Nov 11 |
peter |
57 |
/** |
2648 |
14 Nov 11 |
peter |
\brief Number of data points |
2654 |
15 Nov 11 |
peter |
59 |
|
2654 |
15 Nov 11 |
peter |
\return Number of pairs of data added to object. |
2648 |
14 Nov 11 |
peter |
61 |
*/ |
2648 |
14 Nov 11 |
peter |
62 |
size_t n(void) const; |
2648 |
14 Nov 11 |
peter |
63 |
|
2648 |
14 Nov 11 |
peter |
64 |
/** |
2635 |
09 Nov 11 |
peter |
\brief Lower one-sided p-value |
2654 |
15 Nov 11 |
peter |
66 |
|
2654 |
15 Nov 11 |
peter |
Calculates, \f$ P(R \le r) \f$, the probability to get a score |
2654 |
15 Nov 11 |
peter |
less (or equal) than score() given that there is no association |
2654 |
15 Nov 11 |
peter |
between two variables. |
2654 |
15 Nov 11 |
peter |
70 |
|
2654 |
15 Nov 11 |
peter |
\see p_right() |
2635 |
09 Nov 11 |
peter |
72 |
*/ |
2637 |
10 Nov 11 |
peter |
73 |
double p_left(bool exact=false) const; |
2637 |
10 Nov 11 |
peter |
74 |
|
2635 |
09 Nov 11 |
peter |
75 |
/** |
2635 |
09 Nov 11 |
peter |
\brief Upper one-sided p-value |
2654 |
15 Nov 11 |
peter |
77 |
|
2654 |
15 Nov 11 |
peter |
Calculates, \f$ P(R \ge r) \f$, the probability to get a score |
2654 |
15 Nov 11 |
peter |
greater (or equal) than score() given that there is no |
2654 |
15 Nov 11 |
peter |
association between two variables. |
2654 |
15 Nov 11 |
peter |
81 |
|
2654 |
15 Nov 11 |
peter |
In \a exact mode the score is calculated for each permutation |
2654 |
15 Nov 11 |
peter |
and compared to the actual score. In each permutation one |
2654 |
15 Nov 11 |
peter |
variable, say \a x, is kept fix while the other (\a y) is |
2654 |
15 Nov 11 |
peter |
shuffled in determionsitic fashion. Number of permutations |
2654 |
15 Nov 11 |
peter |
grows quickly as number of data points grows and with no ties |
2654 |
15 Nov 11 |
peter |
there are N! permutations and the exact method gets very |
2654 |
15 Nov 11 |
peter |
expensive. |
2654 |
15 Nov 11 |
peter |
89 |
|
2654 |
15 Nov 11 |
peter |
In approximative mode P-value is calculated using pearson_p_value(). |
2635 |
09 Nov 11 |
peter |
91 |
*/ |
2637 |
10 Nov 11 |
peter |
92 |
double p_right(bool exact=false) const; |
2637 |
10 Nov 11 |
peter |
93 |
|
2635 |
09 Nov 11 |
peter |
94 |
/** |
2635 |
09 Nov 11 |
peter |
\brief Two-sided p-value |
2654 |
15 Nov 11 |
peter |
96 |
|
2654 |
15 Nov 11 |
peter |
Calculates, \f$ P(|R| \ge |r|) \f$, the probability to get a |
2654 |
15 Nov 11 |
peter |
score more extreme (or equal) than score() given that there is |
2654 |
15 Nov 11 |
peter |
no association between two variables. If score(), \a r, is |
2654 |
15 Nov 11 |
peter |
greater than zero, this equals 2 * p_right(). Otherwise it |
2654 |
15 Nov 11 |
peter |
equals 2 * p_left(). |
2635 |
09 Nov 11 |
peter |
102 |
*/ |
2637 |
10 Nov 11 |
peter |
103 |
double p_value(bool exact=false) const; |
2637 |
10 Nov 11 |
peter |
104 |
|
2635 |
09 Nov 11 |
peter |
105 |
/** |
2654 |
15 Nov 11 |
peter |
\brief reset to empty |
2637 |
10 Nov 11 |
peter |
107 |
*/ |
2637 |
10 Nov 11 |
peter |
108 |
void reset(void); |
2637 |
10 Nov 11 |
peter |
109 |
|
2637 |
10 Nov 11 |
peter |
110 |
/** |
2654 |
15 Nov 11 |
peter |
\brief spearman rank correlation coefficient |
2654 |
15 Nov 11 |
peter |
112 |
|
2654 |
15 Nov 11 |
peter |
The Spearman rank correlation coeeficient is defined as \f$ |
2654 |
15 Nov 11 |
peter |
\frac{12\sum (R_i-\frac{n+1}{2})(S_i-\frac{n+1}{2})}{n(n^2-1)} |
2654 |
15 Nov 11 |
peter |
\f$ where \f$ R_i \f$ and \f$ S_i \f$ are ranks of \a X and \a |
2654 |
15 Nov 11 |
peter |
Y, respectively. In case of ties, the rank is defined as the |
2654 |
15 Nov 11 |
peter |
average rank of the ties. |
2635 |
09 Nov 11 |
peter |
118 |
*/ |
2635 |
09 Nov 11 |
peter |
119 |
double score(void) const; |
2635 |
09 Nov 11 |
peter |
120 |
|
2635 |
09 Nov 11 |
peter |
121 |
private: |
2638 |
10 Nov 11 |
peter |
122 |
class Pimpl; |
2638 |
10 Nov 11 |
peter |
123 |
Pimpl* pimpl_; |
2635 |
09 Nov 11 |
peter |
124 |
|
2638 |
10 Nov 11 |
peter |
// Copy not allowed |
2638 |
10 Nov 11 |
peter |
126 |
Spearman(const Spearman& other); |
2638 |
10 Nov 11 |
peter |
127 |
Spearman& operator=(const Spearman&); |
2635 |
09 Nov 11 |
peter |
128 |
}; |
2635 |
09 Nov 11 |
peter |
129 |
|
2635 |
09 Nov 11 |
peter |
130 |
}}} // of namespace statistics, yat, and theplu |
2635 |
09 Nov 11 |
peter |
131 |
#endif |