| T O P I C R E V I E W |
| cosy |
Posted - 10/25/2004 : 05:50:51 AM
Origin Version (Select Help-->About Origin):
Operating System:
Hi,
I wanted to write a program to make a fit to a portion of the input data, first without weight and the subsequent ones with weights.I tried it with the function nag_simple_linear_regression.It works fine without weights buts returns 399 (On entry, at least one of the weights is negative)when the data are weighted.The weights are all positive.
Here is the code:
void slope(string datasetname, int sets,int iter)//datasetname,number of data points to be included and number of iterations
{
Worksheet wksW(datasetname);
Dataset dsA(wksW,0);// Input data X,
Dataset dsB(wksW,1);//Input data Y
Dataset dsC(wksW,2);//X column for the portion of data for fitting
Dataset dsD(wksW,3);//Y column for the portion of data for fitting
Dataset dsE(wksW,4);//weights
dsC.SetSize(sets);//
dsD.SetSize(sets);
dsE.SetSize(sets);
//Nag_SumSquare mean;
int kk;
kk = dsA.GetUpperBound();
printf("Regression of %s\n",datasetname);
printf("Number of data points:%d\n",kk);
char m, w;
int i, n;
double a, b, err_a, err_b, rsq, rss, df;
double *wtptr;
int success;
printf("Starting fit for first polarization cylce without weights\n");
int ii;
int setnum=0;
for(ii=0;ii<sets;ii++)
{
dsC[ii]=dsA[ii+setnum];
dsD[ii]=dsB[ii+setnum];
dsE[ii]=1; // first fit with no weights
}
int doit;
for(doit=0;doit<iter;doit++)
{
success = nag_simple_linear_regression(Nag_AboutMean,sets,dsC,dsD,dsE, &a, &b, &err_a, &err_b,&rsq, &rss, &df);
if(success == 0)
{
printf("\n A ErrA B ErrB RegCoeff SSR DOF\n");
printf("%6.6f\t%6.6f\t%6.6f\t%6.6f\t%6.6f\t%6.6f\t%6.6f\n",a,err_a,b,err_b,rsq,rss,df);
}
else
{
printf("Error %d :The function nag_simple_linear_regression does not work\n",success);
break;
}
printf("Proceeding to weighted fit with weight=1/sigma\n");
double data;
for(ii=0;ii<sets;ii++)
{
data =dsD[ii]-(a+(b*dsC[ii]));
dsE[ii]= 1/sqrt(data*data); // now fit with weights (rugged norm)
}
}
}
========================================================================
and another question about the degrees of freedom :
As far as I have understood it is the number of independent variables needed to describe a system and as a rule of thumb it is the number of data points minus the number of parameters in the fit i.e., if I fit a straight line (A+BX) to a dataset containing 100 parameters the DOF is 98.
But the df value in the function returns a much lower value for the same.
Is there anything wrong with my understanding of DOF?
Thanks,
Deepak Samuel.
Edited by - cosy on 10/25/2004 08:48:14 AM |
| 1 L A T E S T R E P L I E S (Newest First) |
| easwar |
Posted - 10/25/2004 : 10:48:48 PM
Hi Deepak,
You cannot pass dataset objects to a NAG function call. You first need to copy them into vectors and then pass the vectors, such as below:
vector vecC, vecD, vecE;
vecC = dsC;
vecD = dsD;
vecE = dsE;
int doit;
for(doit=0;doit<iter;doit++)
{
success = nag_simple_linear_regression(Nag_AboutMean,sets,vecC,vecD,vecE, &a, &b, &err_a, &err_b,&rsq, &rss, &df);
// etc
The code then works fine, and I get proper DOF values as well. Please try with vectors and post again if you run into more errors.
Easwar
OriginLab
|
|
|
