Simple Multiple Regression in C?

Xeratul · 03-19-2017, 03:39 PM

Hello,

A multiple regression is in math pretty commonly used.
It allows to do things such as :
http://www.nature.com/nmeth/journal/...th.3665-F1.jpg
or also nice
http://www.intechopen.com/source/htm.../image20_w.png

One cool thing is to evaluate the effect of variables on another one.

Would you know, maybe, a simple C programm that would do a multiple regression?

There are few, but let's share the best ones.

thank you

This one is just for x and y.

Code:

    #include <stdio.h>
    #include <math.h>
    int main()
    {
        float xx[5],yy[5],nr=0,dr=0;
        float x[5]={60,61,62,63,65},y[5]={3.1,3.6,3.8,4.0,4.1};
        double sum_y=0,sum_xy=0,sum_x=0,sum_xx=0,sum_x2=0,slope=0,intercept=0,reg;
        int i,n=5;
        for(i=0;i<n;i++)
        {
        xx[i]=x[i]*x[i];
        yy[i]=y[i]*y[i];
        }
        for(i=0;i<n;i++)
        {
        sum_x+=x[i];
        sum_y+=y[i];
        sum_xx+= xx[i];
        sum_xy+= x[i]*y[i];
        }
        nr=(n*sum_xy)-(sum_x*sum_y);
        sum_x2=sum_x*sum_x;
        dr=(n*sum_xx)-sum_x2;
        slope=nr/dr;
        double fac = pow(10, 2);
        slope=round(slope*fac)/fac;       
        intercept=(sum_y- slope*sum_x)/n;
        reg= intercept + (slope*64);
        printf("Slope: %.2f\nIntercept: %.3f\nRegression: %.3f",slope,intercept,reg);
        return 0;
    }

astrogeek · 03-19-2017, 04:54 PM

Although the method may be commonly used and powerful, without the context of a data set of interest, X and Y are only numbers and the graphs are just lines and dots!

Perhaps you could provide a data set of interest to yourself, and explain the relationships among the independent variables and the dependent variable of interest. That would make the problem interesting to others and allow for some context for evaluating the quaity of the code and the resultant predicted values, other than a vague "cool" factor.

Here is a brief, simple introduction to the the idea behind Multiple Regression Analysis for those who may be interested.

When posting topics such as this in the Programming forum, please try to provide a specific question or postulate of interest, and the complete context necessary to understand it, per the Site FAQ and LQ Rules. That helps to maintain a high quality of useful information accessible to future searchers. Vague or general posts with no specific question may be more appropriate for the General forum.

sundialsvcs · 03-19-2017, 07:37 PM

Also, check out the open-source "R" statistics programming language.

It is a statistics package unlike anything you ever saw ... and, today, every mainstream package has an interface to it.

Here are some links:

What makes "R" different is that it is truly a programming language, which just happens to have strong support for statistics and plotting. This is quite different from SAS® or SPSS®, which consist of modules, loaded with options, that are loosely tied together using nothing more than a macro facility. "R" is much smaller, more expressive, faster ... and free.

(Note that "R" generally does things in-memory, which used to be a limitation but not so much anymore.)

Given the easy availability of an open source(!) tool that does so much already, I see little return on investment for rolling-your-own in "C."