1 | #ifndef _theplu_yat_statistics_percentiler_ |
---|
2 | #define _theplu_yat_statistics_percentiler_ |
---|
3 | |
---|
4 | // $Id: Percentiler.h 2076 2009-10-06 22:17:22Z peter $ |
---|
5 | |
---|
6 | /* |
---|
7 | Copyright (C) 2008 Jari Häkkinen, Peter Johansson |
---|
8 | Copyright (C) 2009 Peter Johansson |
---|
9 | |
---|
10 | This file is part of the yat library, http://dev.thep.lu.se/yat |
---|
11 | |
---|
12 | The yat library is free software; you can redistribute it and/or |
---|
13 | modify it under the terms of the GNU General Public License as |
---|
14 | published by the Free Software Foundation; either version 3 of the |
---|
15 | License, or (at your option) any later version. |
---|
16 | |
---|
17 | The yat library is distributed in the hope that it will be useful, |
---|
18 | but WITHOUT ANY WARRANTY; without even the implied warranty of |
---|
19 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
---|
20 | General Public License for more details. |
---|
21 | |
---|
22 | You should have received a copy of the GNU General Public License |
---|
23 | along with yat. If not, see <http://www.gnu.org/licenses/>. |
---|
24 | */ |
---|
25 | |
---|
26 | #include "yat/utility/DataWeight.h" |
---|
27 | #include "yat/utility/iterator_traits.h" |
---|
28 | #include "yat/utility/yat_assert.h" |
---|
29 | #include "yat/utility/WeightIterator.h" |
---|
30 | |
---|
31 | #include <algorithm> |
---|
32 | #include <cassert> |
---|
33 | #include <cmath> |
---|
34 | #include <numeric> |
---|
35 | #include <stdexcept> |
---|
36 | #include <vector> |
---|
37 | |
---|
38 | namespace theplu { |
---|
39 | namespace yat { |
---|
40 | namespace statistics { |
---|
41 | |
---|
42 | /** |
---|
43 | \brief Functor to calculate percentile of a range |
---|
44 | |
---|
45 | \since New in yat 0.5 |
---|
46 | */ |
---|
47 | class Percentiler |
---|
48 | { |
---|
49 | public: |
---|
50 | /** |
---|
51 | \param perc percentile to calculate [0,100]. Default value is |
---|
52 | 50, which implies class will calculate median. |
---|
53 | \param sorted if true class assumes that ranges are already |
---|
54 | sorted, if false the range will copied to a new range which is |
---|
55 | sorted. |
---|
56 | */ |
---|
57 | Percentiler(double perc=50, bool sorted=false); |
---|
58 | |
---|
59 | /** |
---|
60 | Function is a non-mutable function, i.e., \a first and \a last |
---|
61 | can be const_iterators. |
---|
62 | |
---|
63 | The \a Nth percentile is defined such that, for example, when |
---|
64 | having four numbers \f$ 0.69 < 1.41 < 3.14 < 28 \f$, the \a Nth |
---|
65 | percentile is: |
---|
66 | |
---|
67 | - \f$ 0.69 \f$ if \f$ N < 25 \f$ |
---|
68 | - \f$ (0.69+1.41)/2 \f$ if \f$ N=25 \f$ |
---|
69 | - \f$ 1.41 \f$ if \f$ 25 < N < 50 \f$ |
---|
70 | - \f$ (1.41+3.14)/2 \f$ if \f$ N=50 \f$ |
---|
71 | - \f$ 3.14 \f$ if \f$ 50 < N < 75 \f$ |
---|
72 | - \f$ (3.14+28)/2 \f$ if \f$ N=75 \f$ |
---|
73 | - \f$ 28 \f$ if \f$ 75 < N \f$ |
---|
74 | |
---|
75 | Similarily, if we have a weighted range \f$ x_0=0.69, w_0=1 ; |
---|
76 | x_1=1.41, w_1=0 ; x_2=3.14, w_2=0.5 ; x_3=28, w_3=0.5 \f$, we |
---|
77 | calculate the accumulated normalized weight \f$ S_k = \frac |
---|
78 | {100}{\sum w_i}\sum_{i=0}^k w_i \f$ and the percentile is |
---|
79 | |
---|
80 | - \f$ 0.69 \f$ if \f$ N < S_0 \f$ |
---|
81 | - \f$ (0.69+3.14)/2 \f$ if \f$ N=S_0 \f$ |
---|
82 | - \f$ 3.14 \f$ if \f$ S_0 < N < S_2 \f$ |
---|
83 | - \f$ (3.14+28)/2 \f$ if \f$ N=S_2 \f$ |
---|
84 | - \f$ 28 \f$ if \f$ S_2 < N \f$ |
---|
85 | |
---|
86 | Note, that data point with weight zero is completely ignored. |
---|
87 | |
---|
88 | \return percentile of range |
---|
89 | */ |
---|
90 | template<typename RandomAccessIterator> |
---|
91 | double operator()(RandomAccessIterator first, |
---|
92 | RandomAccessIterator last) const |
---|
93 | { |
---|
94 | return calculate(first, last, sorted_, |
---|
95 | typename utility::weighted_iterator_traits<RandomAccessIterator>::type()); |
---|
96 | } |
---|
97 | private: |
---|
98 | double perc_; |
---|
99 | bool sorted_; |
---|
100 | |
---|
101 | // unweighted version |
---|
102 | template<typename RandomAccessIterator> |
---|
103 | double calculate(RandomAccessIterator first, RandomAccessIterator last, |
---|
104 | bool sorted, utility::unweighted_iterator_tag tag) const; |
---|
105 | |
---|
106 | // weighted version |
---|
107 | template<typename RandomAccessIterator> |
---|
108 | double calculate(RandomAccessIterator first, RandomAccessIterator last, |
---|
109 | bool sorted, utility::weighted_iterator_tag tag) const; |
---|
110 | |
---|
111 | // using compiler generated copy |
---|
112 | //Percentiler(const Percentiler&); |
---|
113 | //Percentiler& operator=(const Percentiler&); |
---|
114 | |
---|
115 | }; |
---|
116 | |
---|
117 | // template implementations |
---|
118 | // ///////////////////////// |
---|
119 | |
---|
120 | // unweighted version |
---|
121 | template<typename RandomAccessIterator> |
---|
122 | double |
---|
123 | Percentiler::calculate(RandomAccessIterator first, |
---|
124 | RandomAccessIterator last, |
---|
125 | bool sorted, |
---|
126 | utility::unweighted_iterator_tag tag) const |
---|
127 | { |
---|
128 | // range is one value only is a special case |
---|
129 | if (first+1 == last) |
---|
130 | *first; |
---|
131 | if (sorted) { |
---|
132 | size_t n = last - first; |
---|
133 | // have to take care of this special case |
---|
134 | if (perc_>= 100.0) |
---|
135 | return *(--last); |
---|
136 | if (perc_<= 0.0) |
---|
137 | return *first; |
---|
138 | double j = n * perc_ / 100.0; |
---|
139 | size_t i = static_cast<size_t>(j); |
---|
140 | if (i==j) |
---|
141 | return (first[i]+first[i-1])/2; |
---|
142 | return first[i]; |
---|
143 | } |
---|
144 | |
---|
145 | std::vector<double> v_copy; |
---|
146 | v_copy.reserve(std::distance(first,last)); |
---|
147 | std::copy(first, last, std::back_inserter(v_copy)); |
---|
148 | std::sort(v_copy.begin(), v_copy.end()); |
---|
149 | return calculate(v_copy.begin(), v_copy.end(), true, tag); |
---|
150 | } |
---|
151 | |
---|
152 | |
---|
153 | // weighted version |
---|
154 | template<typename RandomAccessIterator> |
---|
155 | double Percentiler::calculate(RandomAccessIterator first, |
---|
156 | RandomAccessIterator last, |
---|
157 | bool sorted, |
---|
158 | utility::weighted_iterator_tag tag) const |
---|
159 | { |
---|
160 | if (sorted) { |
---|
161 | utility::iterator_traits<RandomAccessIterator> trait; |
---|
162 | std::vector<double> accum_w; |
---|
163 | accum_w.reserve(last-first); |
---|
164 | std::partial_sum(weight_iterator(first), |
---|
165 | weight_iterator(last), |
---|
166 | std::back_inserter(accum_w)); |
---|
167 | |
---|
168 | double w_bound=perc_/100.0*accum_w.back(); |
---|
169 | std::vector<double>::const_iterator upper(accum_w.begin()); |
---|
170 | double margin=1e-10; |
---|
171 | while (*upper <= w_bound+margin && upper!=accum_w.end()) |
---|
172 | ++upper; |
---|
173 | if (upper==accum_w.end()) |
---|
174 | --upper; |
---|
175 | std::vector<double>::const_iterator lower(upper); |
---|
176 | while ( *(lower-1)>=w_bound-margin && lower>accum_w.begin()) |
---|
177 | --lower; |
---|
178 | |
---|
179 | return (trait.data(first+(upper-accum_w.begin()))+ |
---|
180 | trait.data(first+(lower-accum_w.begin())))/2; |
---|
181 | } |
---|
182 | |
---|
183 | std::vector<utility::DataWeight> v_copy; |
---|
184 | v_copy.reserve(last-first); |
---|
185 | std::copy(first, last, std::back_inserter(v_copy)); |
---|
186 | std::sort(v_copy.begin(), v_copy.end()); |
---|
187 | return calculate(v_copy.begin(), v_copy.end(), true, tag); |
---|
188 | } |
---|
189 | |
---|
190 | |
---|
191 | }}} // of namespace statistics, yat, and theplu |
---|
192 | |
---|
193 | #endif |
---|