linear regression

[hawita]

After i have read in a data arranged in two columns, separated by white space characters and comprised of M = 1001 rows(where the data may be considered ordered pairs of points x and y), how do i use linear regression to find the equation of a line that best fits this data : y = mx+b
thanks :)

[ravenous]

Er, is this a C++ problem? Or, just a general maths problem? There are two ways in C++:

1. Write your own functions to do it
2. Use an existing library

I would recommend option 2. I've used the Gnu Scientific Library (GSL) in the past. It's a C library, really, but it is quite comprehensive (and does contain functions for various kinds of linear and non-linear least squares fitting).

Option 1 isn't too hard for a simple y = a + b*x model, but get's more tricky past that, so if you're going to have to fit different models in the future, get into using a library now rather than write your own. Unless your aim is to specifically learn more about fitting algorithms or something.

[hawita]

Thanks for the help. This is what i have so far but i do not think i am on the right track because my program is not even compiling. Please help me .

#include<iostream.h>
#include<fstream.h>

main () {
    float p[2002];
    ifstream fin ("Prog5-Data 00.dat");
    float sumxy = 0;
    float sumx = 0;
    for (int x = 0; x<=1001; x++) {
         fin >> p[x] >> p[y] >>;
         sum = sum + p[x]*p[y]; 
     }
   
     for (int i = 0; i < 1001; i+2) {
         fin >> p[i];
         sumx = sumx + p[i]; 
     }
     float sumy = 0;
     for (int z = 1; z <=1001; z++)
          fin>> p[z];
          sumy = sum + p[z]; 
     }
     
     float sumxx = 0;
     for (int lcv = 0; lcv < 1001; lcv +2) {
          fin >> p[lcv];
          sumxx = sumx + p[lcv]*p[lcv];
      }    
}

[ravenous]

You have a bunch of issues here:

• You should use <iostream> , not <iostream.h> etc.
• You should not use a local array for storing your points in, use a dynamically allocated one, so something like:

  double *points = new double[numberOfPoints];
  
  // Set the points to something
  for( int i = 0; i < numberOfPoints; ++i )
      points[i] = i;
  
  // delete the array when you're done (this is important)
  delete[] points;

• For this kind of application a vector is probably easier and safer:

  #include <vector>
  
  int main()
  {
      std::vector< double > points;
      
      for( int i = 0; i < numberOfPoints; ++i )
          points.push_back(i);
          
      // No need to delete anything (safer!)
      
      return 0;
  }

• Use a separate array/vector for the x and y coordinates (it's clearer):

  std::vector< double > x_points;
  std::vector< double > y_points;
  
  for( int i = 0; i < numberOfPoints; ++i ){
      double x, y;
      fin >> x >> y;
      x_points.push_back(x);
      y_points.push_back(y);
  }

• Don't use a for loop to read from the file (I did it above because I hadn't made this point yet). So, it's better to do something like:

  #include <iostream>
  #include <fstream>
  #include <vector>
  
  int main()
  {
      std::ifstream inFile("filename.txt", std::ios::in);
      if ( inFile.is_open() == false )	return 1;
      
      std::vector< double > x_points;
      std::vector< double > y_points;
      
      while( inFile.fail() == false ){
          double x, y;
          fin >> x >> y;
          x_points.push_back(x);
          y_points.push_back(y);
      }
      
      for( int i = 0; i < x_points.size(); ++i )
          std::cout << x_points[i] << " " << y_points[i] << std::endl;
          
      return 0;
  }

See how you get on with those points :o)