Calculating the probability that a TF1 is a good fit

I would like to know if it is possible to calculate the probability that a TF1 fits a histogram even if you don’t actually perform the fit.

In my case I have calculated the values of my TF1 from a fit to other data however I want to calculate how well the function that I have drawn on the histogram fits the data. At the moment in order to do this I have to unfix one of the parameters, perform the fit, check that the parameter I have allowed to vary hasn’t changed significantly and then print the probability of the fit.

So currently my code looks a bit like this:

for (int i=0; i<12; ++i){ // Getting the results from the other fit gMinuit->GetParameter(i, fitResults[i].mean, fitResults[i].eparab); // Fixing the parameters of my function func1->FixParameter(i, fitResults[i].mean); } // Allowing one of the parameters to vary func1->SetParameter(12, 0.1); // Performing the fit so I can see what the probability is h1->Fit("func1"); cout << "Probability of fit = " << func1->GetProb() << endl;

If I fix all the parameters I get an abnormal termination of the fit. What I would like to do is just fix all the parameters draw my func1 on the histogram and then calculate the probability that this function describes the data. Is this possible?


you can do this by building the chi2 function (the same that is used when fitting).
Attached is an example macro doing it

Best Regards

testChi2FuncTest.C (943 Bytes)