# Double Gaussian Fit + Overestimation of Sigma

Dear All,
I am fitting a distribution which seems to be a sum of two Gaussians with both their means close to zero and simgas different. Unlike \$ROOTSYS/tutorials/multifit.C, the ranges of these two distributions are not different. I tried the following:

``````        //h1 is the total histogram; h1 is the signal and h2 is the background
double par;
TF1 *G1 = new TF1 ("G1","gaus",-3.0,3.0);
TF1 *G2 = new TF1 ("G2","gaus",-3.0,3.0);

//G1->FixParameter(1,0.0);
h2->Fit(G1,"R");
h3->Fit(G2,"R");

G1->GetParameters(&par);
G2->GetParameters(&par);

TF1 *f	= new TF1 ("f","gaus(0)+gaus(3)",-3.0,3.0);
f->SetParameters(par);
h1->Fit(f,"R");
f->SetLineColor(2);

// 6 parameters of the fit: (p0 to p5)
double Ncons  = f->GetParameter(0);
double Gmean= f->GetParameter(1);
double Gwdth = f->GetParameter(2);``````

Usually I see that the fit “f” has very good chi square/ndf and the sigma (p2) is slightly less than each of the distributions. However, occassionally I am getting the sigma to be grossly overestimated ~ sum of individual sigmas of G1 and G2. I found by searching that it might help to fix a certain fit parameter of G1/G2. I tried that as well, but it did not help.

I do not really understand if I am reading off the right parameters. Can anyone please tell me if I am making some mistake? Thanks in advance,
-Regards,
Kolahal

Hi,

There are (just looking at the code) no apparent mistakes in your macro. It can be that sometimes, for certain data, the fit did not converge to the right minimum. Sometimes, when fitting two gaussians, you have always a local minimum, where only one of the gaussian fits the data.
In that case you should re-try the fit with different (possibly closer to the minimum) initial parameters

Best Regards

Lorenzo