1 | #ifndef _theplu_yat_statistics_utility_ |
---|
2 | #define _theplu_yat_statistics_utility_ |
---|
3 | |
---|
4 | // $Id: utility.h 3550 2017-01-03 05:41:02Z peter $ |
---|
5 | |
---|
6 | /* |
---|
7 | Copyright (C) 2004 Jari Häkkinen, Peter Johansson |
---|
8 | Copyright (C) 2005 Peter Johansson |
---|
9 | Copyright (C) 2006 Jari Häkkinen, Peter Johansson, Markus Ringnér |
---|
10 | Copyright (C) 2007, 2008 Jari Häkkinen, Peter Johansson |
---|
11 | Copyright (C) 2009, 2010, 2011, 2013, 2014, 2016 Peter Johansson |
---|
12 | |
---|
13 | This file is part of the yat library, http://dev.thep.lu.se/yat |
---|
14 | |
---|
15 | The yat library is free software; you can redistribute it and/or |
---|
16 | modify it under the terms of the GNU General Public License as |
---|
17 | published by the Free Software Foundation; either version 3 of the |
---|
18 | License, or (at your option) any later version. |
---|
19 | |
---|
20 | The yat library is distributed in the hope that it will be useful, |
---|
21 | but WITHOUT ANY WARRANTY; without even the implied warranty of |
---|
22 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
---|
23 | General Public License for more details. |
---|
24 | |
---|
25 | You should have received a copy of the GNU General Public License |
---|
26 | along with yat. If not, see <http://www.gnu.org/licenses/>. |
---|
27 | */ |
---|
28 | |
---|
29 | #include "Percentiler.h" |
---|
30 | |
---|
31 | #include "yat/classifier/DataLookupWeighted1D.h" |
---|
32 | #include "yat/classifier/Target.h" |
---|
33 | #include "yat/normalizer/utility.h" |
---|
34 | #include "yat/utility/concept_check.h" |
---|
35 | #include "yat/utility/DataIterator.h" |
---|
36 | #include "yat/utility/deprecate.h" |
---|
37 | #include "yat/utility/iterator_traits.h" |
---|
38 | #include "yat/utility/sort_index.h" |
---|
39 | #include "yat/utility/Vector.h" |
---|
40 | #include "yat/utility/VectorBase.h" |
---|
41 | #include "yat/utility/yat_assert.h" |
---|
42 | |
---|
43 | #include <boost/concept_check.hpp> |
---|
44 | #include <boost/iterator/iterator_concepts.hpp> |
---|
45 | #include <boost/iterator/iterator_traits.hpp> |
---|
46 | #include <boost/iterator/permutation_iterator.hpp> |
---|
47 | |
---|
48 | #include <gsl/gsl_math.h> |
---|
49 | #include <gsl/gsl_statistics_double.h> |
---|
50 | |
---|
51 | #include <algorithm> |
---|
52 | #include <cmath> |
---|
53 | #include <iterator> |
---|
54 | #include <stdexcept> |
---|
55 | #include <vector> |
---|
56 | |
---|
57 | namespace theplu { |
---|
58 | namespace yat { |
---|
59 | namespace statistics { |
---|
60 | |
---|
61 | /** |
---|
62 | Adding a range [\a first, \a last) into an object of type T. The |
---|
63 | requirements for the type T is to have an add(double, bool, double) |
---|
64 | function. |
---|
65 | |
---|
66 | Type Requirements: |
---|
67 | - \c ForwardIterator is a \readable_iterator |
---|
68 | - \c ForwardIterator is \single_pass_iterator |
---|
69 | */ |
---|
70 | template <typename T, typename ForwardIterator> |
---|
71 | void add(T& o, ForwardIterator first, ForwardIterator last, |
---|
72 | const classifier::Target& target); |
---|
73 | |
---|
74 | /** |
---|
75 | \brief Benjamini Hochberg multiple test correction |
---|
76 | |
---|
77 | Given a sorted range of p-values such that |
---|
78 | \f$ p_1 \le p_2 \le ... \le p_N \f$ a Benjamnini-Hochberg corrected |
---|
79 | p-value, \c q, is calculated recursively as |
---|
80 | \f$ q_i = \f$ min \f$(p_i \frac{N}{i}, q_{i+1})\f$ with the anchor |
---|
81 | constraint that \f$ q_m = p_m \f$. |
---|
82 | |
---|
83 | Type Requirements: |
---|
84 | - \c BidirectionalIterator1 is a \readable_iterator |
---|
85 | - \c BidirectionalIterator1 is a \bidirectional_traversal_iterator |
---|
86 | - \c BidirectionalIterator1::value type is convertible to \c double |
---|
87 | - \c BidirectionalIterator2 is a \readable_iterator |
---|
88 | - \c BidirectionalIterator2 is a \writable_iterator |
---|
89 | - \c BidirectionalIterator2 is a \bidirectional_traversal_iterator |
---|
90 | - \c double is convertible to \c BidirectionalIterator2::value_type |
---|
91 | |
---|
92 | \since New in yat 0.8 |
---|
93 | */ |
---|
94 | template<typename BidirectionalIterator1, typename BidirectionalIterator2> |
---|
95 | void benjamini_hochberg(BidirectionalIterator1 first, |
---|
96 | BidirectionalIterator1 last, |
---|
97 | BidirectionalIterator2 result); |
---|
98 | |
---|
99 | |
---|
100 | /** |
---|
101 | \brief Benjamini Hochberg multiple test correction |
---|
102 | |
---|
103 | Similar to benjamini_hochberg() but does not assume that input |
---|
104 | range, [first, last), is sorted. The resulting range is the same |
---|
105 | as if sorting input range, call benjamini_hochberg, and unsort |
---|
106 | the result range. |
---|
107 | |
---|
108 | Type Requirements: |
---|
109 | - \c RandomAccessIterator is a \readable_iterator |
---|
110 | - \c RandomAccessIterator is a \random_access_traversal_iterator |
---|
111 | - \c RandomAccessIterator::value type is convertible to \c double |
---|
112 | - \c MutableRandomAccessIterator is a \readable_iterator |
---|
113 | - \c MutableRandomAccessIterator is a \writable_iterator |
---|
114 | - \c MutableRandomAccessIterator is a \random_access_traversal_iterator |
---|
115 | - \c double is convertible to \c MutableRandomAccessIterator::value_type |
---|
116 | |
---|
117 | \since New in yat 0.13 |
---|
118 | */ |
---|
119 | template<typename RandomAccessIterator, typename MutableRandomAccessIterator> |
---|
120 | void benjamini_hochberg_unsorted(RandomAccessIterator first, |
---|
121 | RandomAccessIterator last, |
---|
122 | MutableRandomAccessIterator result); |
---|
123 | |
---|
124 | |
---|
125 | /// |
---|
126 | /// Calculates the probability to get \a k or smaller from a |
---|
127 | /// hypergeometric distribution with parameters \a n1 \a n2 \a |
---|
128 | /// t. Hypergeomtric situation you get in the following situation: |
---|
129 | /// Let there be \a n1 ways for a "good" selection and \a n2 ways |
---|
130 | /// for a "bad" selection out of a total of possibilities. Take \a |
---|
131 | /// t samples without replacement and \a k of those are "good" |
---|
132 | /// samples. \a k will follow a hypergeomtric distribution. |
---|
133 | /// |
---|
134 | /// @return cumulative hypergeomtric distribution functions P(k). |
---|
135 | /// |
---|
136 | double cdf_hypergeometric_P(unsigned int k, unsigned int n1, |
---|
137 | unsigned int n2, unsigned int t); |
---|
138 | |
---|
139 | |
---|
140 | /** |
---|
141 | The entropy is calculated as \f$ - \sum_i p_i \log p_i \f$ where |
---|
142 | \f$p_i = \frac{n_i}{\sum_j n_j} \f$ |
---|
143 | |
---|
144 | Requirements: |
---|
145 | - \c InputIterator must be a \readable_iterator |
---|
146 | - \c InputIterator must be a \single_pass_iterator |
---|
147 | - \c InputIterator::value_type must be convertible to \c double |
---|
148 | |
---|
149 | \since New in yat 0.12 |
---|
150 | */ |
---|
151 | template<typename InputIterator> |
---|
152 | double entropy(InputIterator first, InputIterator last); |
---|
153 | |
---|
154 | /** |
---|
155 | \brief one-sided p-value |
---|
156 | |
---|
157 | This function uses the t-distribution to calculate the one-sided |
---|
158 | p-value. Given that the true correlation is zero (Null |
---|
159 | hypothesis) the estimated correlation, r, after a transformation |
---|
160 | is t-distributed: |
---|
161 | |
---|
162 | \f$ \sqrt{(n-2)} \frac{r}{\sqrt{(1-r^2)}} \in t(n-2) \f$ |
---|
163 | |
---|
164 | \return Probability that correlation is larger than \a r by |
---|
165 | chance when having \a n samples. For \a r larger or equal to 1.0, |
---|
166 | 0.0 is returned. For \a r smaller or equal to -1.0, 1.0 is |
---|
167 | returned. |
---|
168 | */ |
---|
169 | double pearson_p_value(double r, unsigned int n); |
---|
170 | |
---|
171 | /// |
---|
172 | /// @brief Computes the kurtosis of the data in a vector. |
---|
173 | /// |
---|
174 | /// The kurtosis measures how sharply peaked a distribution is, |
---|
175 | /// relative to its width. The kurtosis is normalized to zero for a |
---|
176 | /// gaussian distribution. |
---|
177 | /// |
---|
178 | double kurtosis(const utility::VectorBase&); |
---|
179 | |
---|
180 | |
---|
181 | /** |
---|
182 | \brief Median absolute deviation from median |
---|
183 | |
---|
184 | Function is non-mutable function |
---|
185 | |
---|
186 | Type Requirements: |
---|
187 | - \c RandomAccessIterator must be a \ref |
---|
188 | concept_data_iterator |
---|
189 | - \c RandomAccessIterator must be a \random_access_traversal_iterator |
---|
190 | |
---|
191 | Since 0.6 function also work with a \ref |
---|
192 | concept_weighted_iterator |
---|
193 | */ |
---|
194 | template <class RandomAccessIterator> |
---|
195 | double mad(RandomAccessIterator first, RandomAccessIterator last, |
---|
196 | bool sorted=false); |
---|
197 | |
---|
198 | |
---|
199 | /** |
---|
200 | Median is defined to be value in the middle. If number of values |
---|
201 | is even median is the average of the two middle values. the |
---|
202 | median value is given by p equal to 50. If \a sorted is false |
---|
203 | (default), the range is copied, the copy is sorted, and then |
---|
204 | used to calculate the median. |
---|
205 | |
---|
206 | Function is a non-mutable function, i.e., \a first and \a last |
---|
207 | can be const_iterators. |
---|
208 | |
---|
209 | Type Requirements: |
---|
210 | - \c RandomAccessIterator must be a \ref |
---|
211 | concept_data_iterator |
---|
212 | - \c RandomAccessIterator must be a \random_access_traversal_iterator |
---|
213 | |
---|
214 | \return median of range |
---|
215 | */ |
---|
216 | template <class RandomAccessIterator> |
---|
217 | double median(RandomAccessIterator first, RandomAccessIterator last, |
---|
218 | bool sorted=false); |
---|
219 | |
---|
220 | |
---|
221 | /** |
---|
222 | \brief Calculates the mutual information of \a A. |
---|
223 | |
---|
224 | The elements in A are unnormalized probabilies of the joint |
---|
225 | distribution. |
---|
226 | |
---|
227 | The mutual information is calculated as \f$ \sum \sum p(x,y) \log_2 |
---|
228 | \frac {p(x,y)} {p(x)p(y)} \f$ where |
---|
229 | \f$ p(x,y) = \frac {A_{xy}}{\sum_{x,y} A_{xy}} \f$; |
---|
230 | \f$ p(x) = \sum_y A_{xy} / \sum_{x,y} A_{xy} \f$; |
---|
231 | \f$ p(y) = \sum_x A_{xy} / \sum_{x,y} A_{xy} \f$ |
---|
232 | |
---|
233 | Requirements: |
---|
234 | - \c T must be a model of \ref concept_container_2d |
---|
235 | - \c T::value_type must be convertible to \c double |
---|
236 | |
---|
237 | \return mutual information in bits; if you want in natural base |
---|
238 | multiply with \c M_LN2 (defined in \c gsl/gsl_math.h ) |
---|
239 | |
---|
240 | \since New in yat 0.12 |
---|
241 | */ |
---|
242 | template<class T> |
---|
243 | double mutual_information(const T& A); |
---|
244 | |
---|
245 | /** |
---|
246 | The percentile is determined by the \a p, a number between 0 and |
---|
247 | 100. The percentile is found by interpolation, using the formula |
---|
248 | \f$ percentile = (1 - \delta) x_i + \delta x_{i+1} \f$ where \a |
---|
249 | p is floor\f$((n - 1)p/100)\f$ and \f$ \delta \f$ is \f$ |
---|
250 | (n-1)p/100 - i \f$.Thus the minimum value of the vector is given |
---|
251 | by p equal to zero, the maximum is given by p equal to 100 and |
---|
252 | the median value is given by p equal to 50. If @a sorted |
---|
253 | is false (default), the vector is copied, the copy is sorted, |
---|
254 | and then used to calculate the median. |
---|
255 | |
---|
256 | Function is a non-mutable function, i.e., \a first and \a last |
---|
257 | can be const_iterators. |
---|
258 | |
---|
259 | Requirements: RandomAccessIterator is an iterator over a range of |
---|
260 | doubles (or any type being convertable to double). |
---|
261 | |
---|
262 | @return \a p'th percentile of range |
---|
263 | |
---|
264 | \deprecated percentile2 will replace this function in the future |
---|
265 | |
---|
266 | \note the definition of percentile used here is not identical to |
---|
267 | that one used in percentile2 and Percentile. The difference is |
---|
268 | smaller for large ranges. |
---|
269 | */ |
---|
270 | template <class RandomAccessIterator> |
---|
271 | double percentile(RandomAccessIterator first, RandomAccessIterator last, |
---|
272 | double p, bool sorted=false) YAT_DEPRECATE; |
---|
273 | |
---|
274 | /** |
---|
275 | \see Percentiler |
---|
276 | |
---|
277 | Type Requirements: |
---|
278 | - \c RandomAccessIterator must be a \ref |
---|
279 | concept_data_iterator |
---|
280 | - \c RandomAccessIterator must be a \random_access_traversal_iterator |
---|
281 | |
---|
282 | \since new in yat 0.5 |
---|
283 | */ |
---|
284 | template <class RandomAccessIterator> |
---|
285 | double percentile2(RandomAccessIterator first, RandomAccessIterator last, |
---|
286 | double p, bool sorted=false) |
---|
287 | { |
---|
288 | Percentiler percentiler(p, sorted); |
---|
289 | return percentiler(first, last); |
---|
290 | } |
---|
291 | |
---|
292 | /// |
---|
293 | /// @brief Computes the skewness of the data in a vector. |
---|
294 | /// |
---|
295 | /// The skewness measures the asymmetry of the tails of a |
---|
296 | /// distribution. |
---|
297 | /// |
---|
298 | double skewness(const utility::VectorBase&); |
---|
299 | |
---|
300 | |
---|
301 | //////// template implementations /////////// |
---|
302 | |
---|
303 | template <typename T, typename ForwardIterator> |
---|
304 | void add(T& o, ForwardIterator first, ForwardIterator last, |
---|
305 | const classifier::Target& target) |
---|
306 | { |
---|
307 | BOOST_CONCEPT_ASSERT((boost_concepts::ReadableIterator<ForwardIterator>)); |
---|
308 | BOOST_CONCEPT_ASSERT((boost_concepts::SinglePassIterator<ForwardIterator>)); |
---|
309 | utility::iterator_traits<ForwardIterator> traits; |
---|
310 | for (size_t i=0; first!=last; ++i, ++first) |
---|
311 | o.add(traits.data(first), target.binary(i), traits.weight(first)); |
---|
312 | } |
---|
313 | |
---|
314 | |
---|
315 | template<typename BidirectionalIterator1, typename BidirectionalIterator2> |
---|
316 | void benjamini_hochberg(BidirectionalIterator1 first, |
---|
317 | BidirectionalIterator1 last, |
---|
318 | BidirectionalIterator2 result) |
---|
319 | { |
---|
320 | BOOST_CONCEPT_ASSERT((boost_concepts::ReadableIterator<BidirectionalIterator1>)); |
---|
321 | BOOST_CONCEPT_ASSERT((boost_concepts::BidirectionalTraversal<BidirectionalIterator1>)); |
---|
322 | |
---|
323 | BOOST_CONCEPT_ASSERT((boost_concepts::ReadableIterator<BidirectionalIterator2>)); |
---|
324 | BOOST_CONCEPT_ASSERT((boost_concepts::WritableIterator<BidirectionalIterator2>)); |
---|
325 | BOOST_CONCEPT_ASSERT((boost_concepts::BidirectionalTraversal<BidirectionalIterator2>)); |
---|
326 | |
---|
327 | typedef typename boost::iterator_difference<BidirectionalIterator1>::type |
---|
328 | diff_type; |
---|
329 | |
---|
330 | diff_type n = std::distance(first, last); |
---|
331 | if (!n) |
---|
332 | return; |
---|
333 | std::advance(result, n-1); |
---|
334 | first = last; |
---|
335 | --first; |
---|
336 | diff_type rank = n; |
---|
337 | |
---|
338 | double x = 1.0; |
---|
339 | while (rank) { |
---|
340 | x = std::min(*first * n/static_cast<double>(rank), x); |
---|
341 | *result = x; |
---|
342 | --rank; |
---|
343 | --first; |
---|
344 | --result; |
---|
345 | } |
---|
346 | } |
---|
347 | |
---|
348 | |
---|
349 | template<typename RandomAccessIterator, typename MutableRandomAccessIterator> |
---|
350 | void benjamini_hochberg_unsorted(RandomAccessIterator first, |
---|
351 | RandomAccessIterator last, |
---|
352 | MutableRandomAccessIterator result) |
---|
353 | { |
---|
354 | BOOST_CONCEPT_ASSERT((boost_concepts::ReadableIterator<RandomAccessIterator>)); |
---|
355 | BOOST_CONCEPT_ASSERT((boost_concepts::RandomAccessTraversal<RandomAccessIterator>)); |
---|
356 | |
---|
357 | BOOST_CONCEPT_ASSERT((boost_concepts::ReadableIterator<MutableRandomAccessIterator>)); |
---|
358 | BOOST_CONCEPT_ASSERT((boost_concepts::WritableIterator<MutableRandomAccessIterator>)); |
---|
359 | BOOST_CONCEPT_ASSERT((boost_concepts::RandomAccessTraversal<MutableRandomAccessIterator>)); |
---|
360 | |
---|
361 | std::vector<size_t> idx; |
---|
362 | utility::sort_index(first, last, idx); |
---|
363 | benjamini_hochberg(boost::make_permutation_iterator(first, idx.begin()), |
---|
364 | boost::make_permutation_iterator(first, idx.end()), |
---|
365 | boost::make_permutation_iterator(result, idx.begin())); |
---|
366 | } |
---|
367 | |
---|
368 | |
---|
369 | template<typename InputIterator> |
---|
370 | double entropy(InputIterator first, InputIterator last) |
---|
371 | { |
---|
372 | BOOST_CONCEPT_ASSERT((boost_concepts::ReadableIterator<InputIterator>)); |
---|
373 | BOOST_CONCEPT_ASSERT((boost_concepts::SinglePassIterator<InputIterator>)); |
---|
374 | using boost::Convertible; |
---|
375 | typedef typename boost::iterator_value<InputIterator>::type T; |
---|
376 | BOOST_CONCEPT_ASSERT((Convertible<T,double>)); |
---|
377 | double sum = 0; |
---|
378 | double N = 0; |
---|
379 | for (; first != last; ++first) { |
---|
380 | if (*first) { |
---|
381 | N += *first; |
---|
382 | sum += *first * std::log(static_cast<double>(*first)); |
---|
383 | } |
---|
384 | } |
---|
385 | return -sum / N + log(N); |
---|
386 | } |
---|
387 | |
---|
388 | |
---|
389 | template <class RandomAccessIterator> |
---|
390 | double mad(RandomAccessIterator first, RandomAccessIterator last, |
---|
391 | bool sorted) |
---|
392 | { |
---|
393 | BOOST_CONCEPT_ASSERT((utility::DataIteratorConcept<RandomAccessIterator>)); |
---|
394 | BOOST_CONCEPT_ASSERT((boost_concepts::RandomAccessTraversal<RandomAccessIterator>)); |
---|
395 | double m = median(first, last, sorted); |
---|
396 | typedef typename boost::iterator_value<RandomAccessIterator>::type T; |
---|
397 | std::vector<T> ad(std::distance(first, last)); |
---|
398 | // copy weights if needed |
---|
399 | normalizer::detail::copy_weight_if_weighted(first, last, ad.begin()); |
---|
400 | // assign data (absolute deviation from median) |
---|
401 | utility::iterator_traits<RandomAccessIterator> traits; |
---|
402 | utility::DataIterator<typename std::vector<T>::iterator> first2(ad.begin()); |
---|
403 | while (first!=last) { |
---|
404 | *first2 = std::abs(traits.data(first)-m); |
---|
405 | ++first; |
---|
406 | ++first2; |
---|
407 | } |
---|
408 | return median(ad.begin(), ad.end(), false); |
---|
409 | } |
---|
410 | |
---|
411 | |
---|
412 | template <class RandomAccessIterator> |
---|
413 | double median(RandomAccessIterator first, RandomAccessIterator last, |
---|
414 | bool sorted) |
---|
415 | { |
---|
416 | BOOST_CONCEPT_ASSERT((utility::DataIteratorConcept<RandomAccessIterator>)); |
---|
417 | BOOST_CONCEPT_ASSERT((boost_concepts::RandomAccessTraversal<RandomAccessIterator>)); |
---|
418 | return percentile2(first, last, 50.0, sorted); |
---|
419 | } |
---|
420 | |
---|
421 | |
---|
422 | template<class T> |
---|
423 | double mutual_information(const T& n) |
---|
424 | { |
---|
425 | BOOST_CONCEPT_ASSERT((utility::Container2D<T>)); |
---|
426 | using boost::Convertible; |
---|
427 | BOOST_CONCEPT_ASSERT((Convertible<typename T::value_type,double>)); |
---|
428 | |
---|
429 | // p_x = \sum_y p_xy |
---|
430 | |
---|
431 | // Mutual Information is defined as |
---|
432 | // \sum_xy p_xy * log (p_xy / (p_x p_y)) = |
---|
433 | // \sum_xy p_xy * [log p_xy - log p_x - log p_y] |
---|
434 | // \sum_xy p_xy log p_xy - p_xy log p_x - p_xy log p_y |
---|
435 | // \sum_xy p_xy log p_xy - \sum_x p_x log p_x - \sum_y p_y log p_y |
---|
436 | // - entropy_xy + entropy_x + entropy_y |
---|
437 | |
---|
438 | utility::Vector rowsum(n.columns(), 0); |
---|
439 | for (size_t c = 0; c<n.columns(); ++c) |
---|
440 | rowsum(c) = std::accumulate(n.begin_column(c), n.end_column(c), 0); |
---|
441 | |
---|
442 | utility::Vector colsum(n.rows(), 0); |
---|
443 | for (size_t r = 0; r<n.rows(); ++r) |
---|
444 | colsum(r) = std::accumulate(n.begin_row(r), n.end_row(r), 0); |
---|
445 | |
---|
446 | double mi = - entropy(n.begin(), n.end()); |
---|
447 | mi += entropy(rowsum.begin(), rowsum.end()); |
---|
448 | mi += entropy(colsum.begin(), colsum.end()); |
---|
449 | |
---|
450 | return mi/M_LN2; |
---|
451 | } |
---|
452 | |
---|
453 | |
---|
454 | template <class RandomAccessIterator> |
---|
455 | double percentile(RandomAccessIterator first, RandomAccessIterator last, |
---|
456 | double p, bool sorted) |
---|
457 | { |
---|
458 | BOOST_CONCEPT_ASSERT((boost::RandomAccessIterator<RandomAccessIterator>)); |
---|
459 | BOOST_CONCEPT_ASSERT((utility::DataIteratorConcept<RandomAccessIterator>)); |
---|
460 | // range is one value only is a special case |
---|
461 | if (first+1 == last) |
---|
462 | return utility::iterator_traits<RandomAccessIterator>().data(first); |
---|
463 | if (sorted) { |
---|
464 | // have to take care of this special case |
---|
465 | if (p>=100) |
---|
466 | return utility::iterator_traits<RandomAccessIterator>().data(--last); |
---|
467 | double j = p/100 * (std::distance(first,last)-1); |
---|
468 | int i = static_cast<int>(j); |
---|
469 | return (1-j+floor(j))*first[i] + (j-floor(j))*first[i+1]; |
---|
470 | } |
---|
471 | using std::iterator_traits; |
---|
472 | typedef typename iterator_traits<RandomAccessIterator>::value_type value_t; |
---|
473 | std::vector<value_t> v_copy(first, last); |
---|
474 | size_t i = static_cast<size_t>(p/100 * (v_copy.size()-1)); |
---|
475 | if (i+2 < v_copy.size()) { |
---|
476 | std::partial_sort(v_copy.begin(), v_copy.begin()+i+2, v_copy.end()); |
---|
477 | } |
---|
478 | else |
---|
479 | std::sort(v_copy.begin(), v_copy.end()); |
---|
480 | return percentile(v_copy.begin(), v_copy.end(), p, true); |
---|
481 | } |
---|
482 | |
---|
483 | |
---|
484 | }}} // of namespace statistics, yat, and theplu |
---|
485 | |
---|
486 | #endif |
---|