Normalize THSatck

ROOT
1

Dear expert,
I want to normalize a THStack such that the integral is equal to 1.
For the TH1 I used to do:
h->Scale(1/h->Integral(“width”));
but the the Scale() or even Integral() are not a member of THStack.
Is there a smart way to do that?
Regards

2

A THStack is a collection of histograms, not a single histogram.
What about summing your histograms and doing the normalisation on the result ?

3

Dear Couet,
thank you for your answer. Your advice is interesting, but I would like to see the different contribution from the different bkg, and if I sum all the hist in a single histogram, I won’t be able to see the different bkg (in different color for example like in a THStack). Am I wrong?
Regards

4

Yes you are right.
What about the suggestion given is this old post:
root.cern.ch/root/roottalk/roottalk06/1357.html

5

Dear Couet,
thank you for the link but I’m not sure what he wants to do here.

  • I understood that he want to scale the histo and then add them in a THStack, so at the end I’m not sure how his THStack is normalise.
  • Me, I want to:
    a) put all the histogram in a THStack, and then normalise the THStack area to 1 (but you said it is not possible).
    Or
    b) normalise each histo, such that after after I put them in the THStack, the THStack area is 1 (but I’m not sure this makes sense at this point…).
6

Yes you could do that but have to know to which value each histo should be normalized so the sum will be 1 …

7

You could also try something like this:

// assuming that "MyTHStack" contains TH1F histograms ...
TH1F *h = new TH1F(*((TH1F *)(MyTHStack->GetStack()->Last())));
h->SetNameTitle("h", "my normalized stack");
h->Scale(1./h->Integral("width"));
8

Dear Pepe,
if I normalize independently each TH1 in the stack to 1, and then add them to the stack, the THStack area will not be >1, am I wrong? What I want to do is to add my differents bkg to the THStack and then normalize the THStack area to 1 (or normalize each histo of the hstack such that the THStack area is egal to 1, but I do not want to resolve complex equation…). Do you see a smart way to do that?
Regards

9

If you just need the “sum histogram”, use my trick (you will get a normalized “copy”).
Otherwise, you will need to scale all “N” contributing histograms independently (for example, in the simplest case, to a value “1./N”).

10

Dear Pepe,
ok, thank you for your answer.
Regards