Residual plot fitting

Newbie
1

Hii ,
I have a data set ( text file) . I plot a graph and fitted it using quadratic equation.
I need to fit the residuals of fitting. I can do it for histograms but how do I do it for graph ?
Can you privide some code ?

Dummy text file :

Energy     Sigma   Sigmaerror

439           1.4           .14
450           1.2            .04
583           1.4            .05
610            1.4            .04
1730         1.5             .09
1765         1.6            .04
2050         1.9             .13
2205         1.9            .08
2614         1.9             .05
2904         1.9             .07
4082          2.6            .17
5551         3.7               .19
6475          4.01           .27
6986         3.99            .24

I fitted this using a0+a1x+a2x**2 where x is energy

Thanking you,

Saradindu

2

I guess @moneta can help.
It would be nice to post the macro you have now.

3

Hello ,
This is used to plot the data file . After that I just used fit function of root to fit the data points . I don’t have macro for plotting the residuals .

Int_t n=100;
Float_t x1,ex1,y1,ey1;
Int_t i=0;
Float_t x[100],ex[100],y[100],ey[100];
while(1){
        ncols=fscanf(fp,"%f %f %f",&x1,&y1,&ey1);
        if (ncols<0) break;
        x[i]=x1;
//        ex[i]=ex1;
        y[i]=y1;
        ey[i]=ey1;
//        cout<<x[i]<<"\t"<<ex[i]<<"\t"<<y[i]<<"\t"<<ey[i]<<endl;
        cout<<x1<<"\t"<<ex1<<"\t"<<y1<<"\t"<<ey1<<endl;
        i++;
        nlines++;
        }
cout<<"found"<<nlines<<endl;
fclose(fp);

TGraphErrors*gr=new TGraphErrors(nlines,x,y,ex,ey);
TCanvas*c1=new TCanvas("c1","graph",100,10,900,600);
c1->SetFillColor(0);

gr->SetMaximum(6);
gr->SetMinimum(0);

gr->SetMarkerStyle(21);
gr->SetLineColor(1);
gr->SetTitle(" Sigma vs Energy Graph");
gr->SetMarkerSize(1);
gr->GetXaxis()->CenterTitle();
gr->GetYaxis()->CenterTitle();
gr->GetYaxis()->SetTitle(" Sigma (keV)");
gr->GetXaxis()->SetTitle(" Energy (keV)");
gr->Draw("AP");
return c1;
}

Thanks
Saradindu

4

Hi,
If you have an histogram you can use TRatioPlot. For a Graph, you can early loop on the data points and compute the residuals of the fit and fill a new TGraphErrors.
If you don’t know how to do it, I could make a small example for you

Cheers

Lorenzo

5

I think I have one ready:

{
  TRandom* rand = new TRandom;

  TGraphErrors* gra = new TGraphErrors;
  for(Int_t i = 1; i <= 10; i++){
    Double_t x = i*100;
    Double_t y = rand->Gaus(x, TMath::Sqrt(x));
    gra->SetPoint(i - 1, x, y);
    gra->SetPointError(i - 1, 0, TMath::Sqrt(y));
  } // i

  TCanvas* can = new TCanvas("can");
  can->SetBottomMargin(0.3);
  //  gra->Draw("ape");
  TF1* f1 = new TF1("f1", "[0]*x");
  gra->Fit(f1);
  gra->Draw("ape");
  f1->Draw("l same");
  
  TPad* pad = new TPad("pad", "pad", 0., 0., 1., 1.);
  pad->SetTopMargin(0.7);
  pad->Draw();
  pad->SetFillStyle(0);
  pad->cd();

  TGraphErrors* residual = new TGraphErrors;
  for(Int_t i = 1; i <= 10; i++){
    Double_t x = i*100;
    Double_t y0 = f1->Eval(x);
    Double_t y = (gra->GetY()[i - 1] - y0)/TMath::Sqrt(y0);
    Double_t ey = gra->GetErrorY(i - 1)/TMath::Sqrt(y0);
    residual->SetPoint(i - 1, x, y);
    residual->SetPointError(i - 1, 0, ey);
  } // i
  
  residual->GetHistogram()->SetMaximum(5);
  residual->GetHistogram()->SetMinimum(-5);
  residual->Draw("ape");
}

Screenshot 2021-11-18 at 16.16.39
Screenshot 2021-11-18 at 16.16.391286×1682 112 KB

6

Thank you very much . I am able to do it for histogram .
But I could not do it for Graph .

If you put one example using graph it will be very much helpful .

Regards
Saradindu

7

Thank you very much .
I am trying with this and get back to you.

regards
Saradindu

1 Like