4255 |
16 Dec 22 |
peter |
1 |
#ifndef _theplu_yat_statistics_kaplan_meier |
4255 |
16 Dec 22 |
peter |
2 |
#define _theplu_yat_statistics_kaplan_meier |
4255 |
16 Dec 22 |
peter |
3 |
|
4255 |
16 Dec 22 |
peter |
// $Id$ |
4255 |
16 Dec 22 |
peter |
5 |
|
4255 |
16 Dec 22 |
peter |
6 |
/* |
4255 |
16 Dec 22 |
peter |
Copyright (C) 2022 Peter Johansson |
4255 |
16 Dec 22 |
peter |
8 |
|
4255 |
16 Dec 22 |
peter |
This file is part of the yat library, https://dev.thep.lu.se/yat |
4255 |
16 Dec 22 |
peter |
10 |
|
4255 |
16 Dec 22 |
peter |
The yat library is free software; you can redistribute it and/or |
4255 |
16 Dec 22 |
peter |
modify it under the terms of the GNU General Public License as |
4255 |
16 Dec 22 |
peter |
published by the Free Software Foundation; either version 3 of the |
4255 |
16 Dec 22 |
peter |
License, or (at your option) any later version. |
4255 |
16 Dec 22 |
peter |
15 |
|
4255 |
16 Dec 22 |
peter |
The yat library is distributed in the hope that it will be useful, |
4255 |
16 Dec 22 |
peter |
but WITHOUT ANY WARRANTY; without even the implied warranty of |
4255 |
16 Dec 22 |
peter |
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
4255 |
16 Dec 22 |
peter |
General Public License for more details. |
4255 |
16 Dec 22 |
peter |
20 |
|
4255 |
16 Dec 22 |
peter |
You should have received a copy of the GNU General Public License |
4255 |
16 Dec 22 |
peter |
along with yat. If not, see <https://www.gnu.org/licenses/>. |
4255 |
16 Dec 22 |
peter |
23 |
*/ |
4255 |
16 Dec 22 |
peter |
24 |
|
4255 |
16 Dec 22 |
peter |
25 |
#include <map> |
4255 |
16 Dec 22 |
peter |
26 |
|
4255 |
16 Dec 22 |
peter |
27 |
namespace theplu { |
4255 |
16 Dec 22 |
peter |
28 |
namespace yat { |
4255 |
16 Dec 22 |
peter |
29 |
namespace statistics { |
4255 |
16 Dec 22 |
peter |
30 |
|
4255 |
16 Dec 22 |
peter |
31 |
/** |
4255 |
16 Dec 22 |
peter |
\brief Kaplan-Meier estimator |
4255 |
16 Dec 22 |
peter |
33 |
|
4255 |
16 Dec 22 |
peter |
Data is evaluated lazily, i.e., data is not computed when added |
4255 |
16 Dec 22 |
peter |
but when data is accessed through the functions returning |
4255 |
16 Dec 22 |
peter |
iterators to the underlying data structure. Since computing teh |
4255 |
16 Dec 22 |
peter |
data is linear in number of time points, the typical use case is |
4255 |
16 Dec 22 |
peter |
to first add all the data and then access the estimated values. |
4255 |
16 Dec 22 |
peter |
39 |
|
4255 |
16 Dec 22 |
peter |
\since New in yat 0.21 |
4255 |
16 Dec 22 |
peter |
41 |
*/ |
4255 |
16 Dec 22 |
peter |
42 |
class KaplanMeier |
4255 |
16 Dec 22 |
peter |
43 |
{ |
4255 |
16 Dec 22 |
peter |
44 |
public: |
4255 |
16 Dec 22 |
peter |
45 |
/** |
4255 |
16 Dec 22 |
peter |
Class representing a time point in a kaplan-meier estimator |
4255 |
16 Dec 22 |
peter |
47 |
*/ |
4255 |
16 Dec 22 |
peter |
48 |
class TimePoint |
4255 |
16 Dec 22 |
peter |
49 |
{ |
4255 |
16 Dec 22 |
peter |
50 |
public: |
4255 |
16 Dec 22 |
peter |
51 |
/** |
4255 |
16 Dec 22 |
peter |
Default constructor |
4255 |
16 Dec 22 |
peter |
53 |
*/ |
4255 |
16 Dec 22 |
peter |
54 |
TimePoint(void); |
4255 |
16 Dec 22 |
peter |
55 |
|
4255 |
16 Dec 22 |
peter |
56 |
/** |
4255 |
16 Dec 22 |
peter |
\param event number of events occuring at this time point |
4255 |
16 Dec 22 |
peter |
\param censored number of data points being censored at this |
4255 |
16 Dec 22 |
peter |
time point. |
4255 |
16 Dec 22 |
peter |
60 |
*/ |
4255 |
16 Dec 22 |
peter |
61 |
TimePoint(unsigned int event, unsigned int censored); |
4255 |
16 Dec 22 |
peter |
62 |
|
4255 |
16 Dec 22 |
peter |
63 |
/** |
4255 |
16 Dec 22 |
peter |
Number of subjects being at risk at this point in time, i.e., |
4255 |
16 Dec 22 |
peter |
number of data points that occur or are being censored at |
4255 |
16 Dec 22 |
peter |
this time point or later. |
4255 |
16 Dec 22 |
peter |
67 |
*/ |
4255 |
16 Dec 22 |
peter |
68 |
unsigned int at_risk(void) const; |
4255 |
16 Dec 22 |
peter |
69 |
|
4255 |
16 Dec 22 |
peter |
70 |
/** |
4255 |
16 Dec 22 |
peter |
Number of data points being censored at this time point |
4255 |
16 Dec 22 |
peter |
72 |
*/ |
4255 |
16 Dec 22 |
peter |
73 |
unsigned int censored(void) const; |
4255 |
16 Dec 22 |
peter |
74 |
|
4255 |
16 Dec 22 |
peter |
75 |
/** |
4255 |
16 Dec 22 |
peter |
Number of events occurring |
4255 |
16 Dec 22 |
peter |
77 |
*/ |
4255 |
16 Dec 22 |
peter |
78 |
unsigned int event(void) const; |
4255 |
16 Dec 22 |
peter |
79 |
|
4255 |
16 Dec 22 |
peter |
80 |
/** |
4255 |
16 Dec 22 |
peter |
\return an estimate of proportions of samples alive. |
4255 |
16 Dec 22 |
peter |
82 |
|
4255 |
16 Dec 22 |
peter |
The proportion is estimated as \f$ s(t) \prod_{i:t_i\le t} |
4255 |
16 Dec 22 |
peter |
\left( 1 - \frac{x_i}{n_i} \right) \f$, where \f$d_i\f$ is |
4255 |
16 Dec 22 |
peter |
number of events at time \f$t_i\f$ and \f$ n_i \f$ is subjects |
4255 |
16 Dec 22 |
peter |
at risk at time \f$ t_i\f$ (see at_risk()). |
4255 |
16 Dec 22 |
peter |
87 |
*/ |
4255 |
16 Dec 22 |
peter |
88 |
double survival_proportion(void) const; |
4255 |
16 Dec 22 |
peter |
89 |
private: |
4255 |
16 Dec 22 |
peter |
90 |
friend class KaplanMeier; |
4255 |
16 Dec 22 |
peter |
91 |
unsigned int censored_; |
4255 |
16 Dec 22 |
peter |
92 |
unsigned int event_; |
4255 |
16 Dec 22 |
peter |
93 |
unsigned int n_; |
4255 |
16 Dec 22 |
peter |
94 |
double s_; |
4255 |
16 Dec 22 |
peter |
95 |
}; |
4255 |
16 Dec 22 |
peter |
96 |
|
4255 |
16 Dec 22 |
peter |
97 |
/** |
4255 |
16 Dec 22 |
peter |
Const iterator used to iterate through the underlying data |
4255 |
16 Dec 22 |
peter |
structure. |
4255 |
16 Dec 22 |
peter |
100 |
*/ |
4255 |
16 Dec 22 |
peter |
101 |
typedef std::map<double, TimePoint>::const_iterator const_iterator; |
4255 |
16 Dec 22 |
peter |
102 |
|
4255 |
16 Dec 22 |
peter |
103 |
/** |
4255 |
16 Dec 22 |
peter |
Default constructor |
4255 |
16 Dec 22 |
peter |
105 |
*/ |
4255 |
16 Dec 22 |
peter |
106 |
KaplanMeier(void); |
4255 |
16 Dec 22 |
peter |
107 |
|
4255 |
16 Dec 22 |
peter |
108 |
/** |
4255 |
16 Dec 22 |
peter |
Add a data point at time \c time. If \c event is true, counted |
4255 |
16 Dec 22 |
peter |
as an event (e.g. death); otherwise counted as sample being |
4255 |
16 Dec 22 |
peter |
censored at time \c time. |
4255 |
16 Dec 22 |
peter |
112 |
|
4255 |
16 Dec 22 |
peter |
Takes logarithmic time in terms of already added time points. |
4255 |
16 Dec 22 |
peter |
114 |
*/ |
4255 |
16 Dec 22 |
peter |
115 |
void add(double time, bool event); |
4255 |
16 Dec 22 |
peter |
116 |
|
4255 |
16 Dec 22 |
peter |
117 |
/** |
4255 |
16 Dec 22 |
peter |
Add \c event events and \c censored censored data points at |
4255 |
16 Dec 22 |
peter |
time \c time. |
4255 |
16 Dec 22 |
peter |
120 |
|
4255 |
16 Dec 22 |
peter |
Takes logarithmic time in terms of already added time points. |
4255 |
16 Dec 22 |
peter |
122 |
*/ |
4255 |
16 Dec 22 |
peter |
123 |
void add(double time, unsigned int event, unsigned int censored); |
4255 |
16 Dec 22 |
peter |
124 |
|
4255 |
16 Dec 22 |
peter |
125 |
/** |
4255 |
16 Dec 22 |
peter |
\return an iterator to the earliest data point with data |
4255 |
16 Dec 22 |
peter |
127 |
*/ |
4255 |
16 Dec 22 |
peter |
128 |
const_iterator begin(void) const; |
4255 |
16 Dec 22 |
peter |
129 |
|
4255 |
16 Dec 22 |
peter |
130 |
/** |
4255 |
16 Dec 22 |
peter |
Clear all data. |
4255 |
16 Dec 22 |
peter |
132 |
*/ |
4255 |
16 Dec 22 |
peter |
133 |
void clear(void); |
4255 |
16 Dec 22 |
peter |
134 |
|
4255 |
16 Dec 22 |
peter |
135 |
/** |
4255 |
16 Dec 22 |
peter |
\return past last iterator |
4255 |
16 Dec 22 |
peter |
137 |
*/ |
4255 |
16 Dec 22 |
peter |
138 |
const_iterator end(void) const; |
4255 |
16 Dec 22 |
peter |
139 |
|
4255 |
16 Dec 22 |
peter |
140 |
/** |
4255 |
16 Dec 22 |
peter |
\return If there is any data at time \c time, return an |
4255 |
16 Dec 22 |
peter |
iterator pointing to it. Otherwise, end() is returned. |
4255 |
16 Dec 22 |
peter |
143 |
*/ |
4255 |
16 Dec 22 |
peter |
144 |
const_iterator find(double time) const; |
4255 |
16 Dec 22 |
peter |
145 |
|
4255 |
16 Dec 22 |
peter |
146 |
/** |
4255 |
16 Dec 22 |
peter |
\return first time point that is not earlier than \c time |
4255 |
16 Dec 22 |
peter |
148 |
*/ |
4255 |
16 Dec 22 |
peter |
149 |
const_iterator lower_bound(double time) const; |
4255 |
16 Dec 22 |
peter |
150 |
|
4255 |
16 Dec 22 |
peter |
151 |
/** |
4255 |
16 Dec 22 |
peter |
\return first time point that is later than \c time |
4255 |
16 Dec 22 |
peter |
153 |
*/ |
4255 |
16 Dec 22 |
peter |
154 |
const_iterator upper_bound(double time) const; |
4255 |
16 Dec 22 |
peter |
155 |
private: |
4255 |
16 Dec 22 |
peter |
// Calculated data in data_ is not calculated every time data is |
4255 |
16 Dec 22 |
peter |
// added, but calculated lazily first when needed, i.e., when user |
4255 |
16 Dec 22 |
peter |
// is accessing the data. |
4255 |
16 Dec 22 |
peter |
159 |
mutable bool updated_; |
4255 |
16 Dec 22 |
peter |
160 |
mutable std::map<double, TimePoint> data_; |
4255 |
16 Dec 22 |
peter |
161 |
void update(void) const; |
4255 |
16 Dec 22 |
peter |
162 |
unsigned int n_; |
4255 |
16 Dec 22 |
peter |
163 |
}; |
4255 |
16 Dec 22 |
peter |
164 |
|
4255 |
16 Dec 22 |
peter |
165 |
}}} // of namespace statistics, yat, and theplu |
4255 |
16 Dec 22 |
peter |
166 |
|
4255 |
16 Dec 22 |
peter |
167 |
#endif |