# source:trunk/test/regression_test.cc@430

Last change on this file since 430 was 430, checked in by Peter, 17 years ago

changed interface of Regression::Local

• Property svn:eol-style set to `native`
• Property svn:keywords set to `Id`
File size: 3.6 KB
Line
1// \$Id: regression_test.cc 430 2005-12-08 22:53:08Z peter \$
2
3
4#include <c++_tools/gslapi/matrix.h>
5#include <c++_tools/statistics/KernelBox.h>
6#include <c++_tools/statistics/Linear.h>
7#include <c++_tools/statistics/LinearWeighted.h>
8#include <c++_tools/statistics/Local.h>
9#include <c++_tools/statistics/Naive.h>
10#include <c++_tools/statistics/NaiveWeighted.h>
11#include <c++_tools/statistics/Polynomial.h>
12
13#include <cmath>
14
15#include <fstream>
16#include <iostream>
17
18
19using namespace theplu;
20
21bool Local_test(statistics::regression::OneDimensionalWeighted&,
22                statistics::regression::Kernel&);
23
24
25int main(const int argc,const char* argv[])
26{
27  using namespace theplu;
28  std::ostream* error;
29  if (argc>1 && argv[1]==std::string("-v"))
30    error = &std::cerr;
31  else {
32    error = new std::ofstream("/dev/null");
33    if (argc>1)
34      std::cout << "regression_test -v : for printing extra information\n";
35  }
36  *error << "testing regression" << std::endl;
37  bool ok = true;
38
39  // test data for Linear and Naive (Weighted and non-weighted)
40  gslapi::vector x(4); x(0)=1970; x(1)=1980; x(2)=1990; x(3)=2000;
41  gslapi::vector y(4); y(0)=12;   y(1)=11;   y(2)=14;   y(3)=13;
42  gslapi::vector w(4); w(0)=0.1;  w(1)=0.2;  w(2)=0.3;  w(3)=0.4;
43
44  *error << "testing regression::LinearWeighted" << std::endl;
45  statistics::regression::LinearWeighted linear_w;
46  linear_w.fit(x,y,w);
47  double y_predicted=0;
48  double y_predicted_err=0;
49  linear_w.predict(1990,y_predicted,y_predicted_err);
50  if (y_predicted!=12.8){
51    *error << "regression_LinearWeighted: cannot reproduce fit." << std::endl;
52    ok=false;
53  }
54
55  // testing regression::NaiveWeighted
56  *error << "testing regression::NaiveWeighted" << std::endl;
57  statistics::regression::NaiveWeighted naive_w;
58  naive_w.fit(x,y,w);
59  y_predicted=0;
60  y_predicted_err=0;
61  naive_w.predict(0.0,y_predicted,y_predicted_err);
62  if (y_predicted!=(0.1*12+0.2*11+0.3*14+0.4*13)) {
63    *error << "regression_NaiveWeighted: cannot reproduce fit." << std::endl;
64    ok=false;
65  }
66
67  // testing regression::Local
68  *error << "testing regression::Local" << std::endl;
69  statistics::regression::KernelBox kb;
70  statistics::regression::LinearWeighted rl;
71  if (!Local_test(rl,kb)) {
72    *error << "regression_Local: Linear cannot reproduce fit." << std::endl;
73    ok=false;
74  }
75  statistics::regression::NaiveWeighted rn;
76  if (!Local_test(rn,kb)) {
77    *error << "regression_Local: Naive cannot reproduce fit." << std::endl;
78    ok=false;
79  }
80
81  // testing regression::Polynomial
82  *error << "testing regression::Polynomial" << std::endl;
83  {
84    std::ifstream s("data/regression_gauss.data");
85    gslapi::matrix data(s);
86    gslapi::vector x(data.rows());
87    gslapi::vector ln_y(data.rows());
88    for (size_t i=0; i<data.rows(); ++i) {
89      x(i)=data(i,0);
90      ln_y(i)=log(data(i,1));
91    }
92
93    statistics::regression::Polynomial polynomialfit(2);
94    polynomialfit.fit(x,ln_y);
95    gslapi::vector fit=polynomialfit.fit_parameters();
96    if (fabs(fit[0]-1.012229646706 + fit[1]-0.012561322528 +
97             fit[2]+1.159674470130)>1e-11) {  // Jari, fix number!
98      *error << "regression_Polynomial: cannot reproduce fit." << std::endl;
99      ok=false;
100    }
101  }
102
103  *error << "testing regression::Linear" << std::endl;
104  statistics::regression::Linear lin;
105
106  *error << "testing regression::Naive" << std::endl;
107  statistics::regression::Naive naive;
108
109  if (error!=&std::cerr)
110    delete error;
111
112  return (ok ? 0 : -1);
113}
114
115
116
117bool Local_test(statistics::regression::OneDimensionalWeighted& r,
118                statistics::regression::Kernel& k)
119{
120  statistics::regression::Local rl(r,k);
121  gslapi::vector points(1000);
122  for (size_t i=0; i<1000; i++){