Problem fitting a function in one range, and plotting in a different range

ROOT
1

Hi,

I have a 1-d histogram that I wish to fit with a function (TF1). The histogram runs from 500 MeV/c^2 to 2.2 GeV^2. I would like to fit the function in a limited range of the data (and get the parameters and chi^2 for that range), but then draw the function over the entire range of the histogram, to visually see the extrapolation. I can restrict the fit range with “R”, but then I have not figured out how to draw it over the entire histogram range. I also tried cloning or copying the function, figuring I could then extend the range, but couldn’t get that to work either.

What is the easiest way to do this?

Spencer Klein
LBNL

2

Try this: fit first the function without drawing. (Option 0 or sth like that).
Then call f->SetRange(min,max); f->Draw("same");

3

Doh - thank you very much.

Now I’m having a different problem. I want to expand my fitting function from 11 to 14 fitting parameters:

TF1 *func = new TF1(“func”,fitfBW1,0.6,2.2,14);
func->SetParameters(0.767,0.151,700.,-500.,1.E5, 1E5,24.,0.782,0.1,1.2 ,0.,1650.,130.,1.);

The TF1 seems to define OK, but whenever I try to use the function, I get error messages like

Users/sklein/STARstuff/twoprong/./dipionBW1.C:153:7: error: no matching member
function for call to 'SetParameters’
func->SetParameters(0.767,0.151,700.,-500.,1.E5,-1E5,24.,0.782,0.1,1.2,0…
^~~~~~~
/opt/local/libexec/root6/include/root/TF1.h:443:21: note: candidate function not
viable: requires at most 11 arguments, but 14 were provided
virtual void SetParameters(Double_t p0,Double_t p1,Double_t p2=0,…

with a similar error for

func->SetParNames("…

Am I missing something, or is this a fundamental limitation of ROOT? I am aware how difficult it is to fit to a function with 14 parameters, but it is easier for me to fix, if needed some parameters with func->FixParameter(10,0.), rather than having to hard-code masses, etc. into my fit function. I could not find documentation on the maximum number of parameters in a TF1; if this exists, it would be nice to document it.

Thanks for your help.

Spencer Klein

4

Yes, there is a limitation for just 11 parameters if given as function arguments. Try instead to pass the arguments as an array:

Double_t pars[14] = {0.767,0.151,700.,-500.,1.E5, 1E5,24.,0.782,0.1,1.2 ,0.,1650.,130.,1.}; func->SetParameters(pars);

5

Thanks! It would be nice to update the documentation to note the limitation on the number of parameters. Also, I was unable to get the argument-as-array approach to work for SetParNames, but can set each name individually with SetParName.

One more fit question: I have a fairly complicated TH1F which includes bins with different widths, read in from a file. I want to fit this function, and then histogram the residuals by subtracting the fit function from the input histogram. This seems, in principle fairly straightforward:

TH1F resid = (TH1F)massalldiff->Clone(“resid”);

for (int j=1;j<numbins;j++) {
double midbin=resid->GetBinCenter(j);
double differ = massalldiff->GetBinContent(j) - func->Eval(midbin);
resid->SetBinContent(j,differ);
}
resid->Draw();

Here massalldiff is the histogram, and func is the function.

and I am pretty sure that something like this worked in ROOT 5. However, I cannot get it to work properly in ROOT6.06, because when I clone the histogram, it also brings along the fit function; when I draw the resid, in addition to the residuals, I also get the fit function, which I both don’t want, and which messes up the scale. I tried 'resid->Draw(“0”), but that didn’t work, and I also tried cloning the histogram before the fit, but that didn’t work either. I would very much rather not have to create the new histogram by copying the new bin widths, etc. element by element from the old histogram. Any ideas?

Spencer

6

Well, here it is more or less obvious:
root.cern.ch/doc/master/classTF … 29c59c9828

Yes, consistency/coherence is not a strength of the ROOT interfaces… :wink:
You could suggest the developers to implement for ROOT6 the function SetParNames(array of names).

[quote=“spencer”]I would very much rather not have to create the new histogram by copying the new bin widths, etc. element by element from the old histogram. Any ideas?
[/quote]
Option a)

resid->GetListOfFunctions()->Clear();
resid->Draw();

Option b)

massaldiff->Fit(func,"N0");
TH1F *resid = (TH1F*)massalldiff->Clone("resid");
resid->Draw();

See:
root.cern.ch/root/html532/TH1.html#TH1:Fit