Nick van Eijndhoven wrote: > > Dear ROOTers, > I send this mail just to have a clear idea how exactly > the errors are treated when adding etc.... several histograms. > I have two 1-D histograms (TH1F *h1, *h2) for which I have set > the error for each bin by hand via the SetBinError() facility. > This I had to do because h1 and h2 are particle spectra which have > been corrected for reconstruction efficiencies and obviously these > rec. eff. factors have also an uncertainty. > Now I have to sum these h1 and h2 into a new histogram : > > TH1F *h3->Add(h1,h2) > > Now my question is whether the error of each bin in h3 is correctly > set to the sum of the errors of h1 and h2 bins or not. > Or is the error of the h3 bin just set to the sqrt() of the bin > content ? > Note that I did NOT specify h3->Sumw2() since from the docs I > understand that in that case the error of each h3 bin is just > the sqrt() of the sum of the squared errors of the h1 and h2 bins. > The latter is correct for multiplication and division of independent > histos, but not for additions and subtractions. > Or do I misunderstand something here ? > Of course I could calculate and set all the errors by hand, but > it is convenient if TH1::Add() would treat the errors correctly. > It would make sense to me if h3->Sumw2() would trigger the > correct error handling in case of TH1::Add(), whereas without the > h3->Sumw2() the error would become just the sqrt() of the bin content. > -- Nick, To fill an histogram h3 with an operation on h1 and h2 with correct errors calculations, one possible sequence is: //Create h3 as a clone of h1 or h2 TH1F *h3 = (TH1F*)h1->Clone(); //Make sure Sumw2 is called for h3 h3->Sumw2(); //Perform operation h3->Add(h1,h2,c1,c2); Rene
This archive was generated by hypermail 2b29 : Tue Jan 04 2000 - 00:34:33 MET