Scaling of histogram sets BinError to 0

[philtk74]

Dear all,

I see very strange behaviour for a histogram: when scaling with 1.0 via the function Scale(), the errors of all bins are set to 0.0 afterwards. I attached a little script I wrote and the corresponding input root file. The corresponding output in my case is:

Integral is 92680.95672607422
Bin content 5539.64990234375, bin error 74.42882440522455
Scale histogram with 1.0
Integral is 92680.95672607422
Bin content 5539.64990234375, bin error 0.0

Any help is very much appreciated!

tmp_renamed.root (120.8 KB)

ScaleBug.py (632 Bytes)

---

Please read tips for efficient and successful posting and posting code

_ROOT Version: 6.26/08
_Platform: CentOS7
Compiler: Not Provided

---

[couet]

Maybe that’s connected to the “IMPORTANT NOTE” here: ROOT: TH1 Class Reference

@moneta can help I am sure.

[moneta]

Hi,

This is caused by the fact that the entries of the histograms are zero. I don’t know why, since it is coming from a file, maybe we should remove this check in TH1::Sumw2 that is used to fill the histogram bin errors correctly.
As a workaround just do before scaling:

memory.SetEntries(memory.GetSumOfWeights())

Lorenzo

[philtk74]

Hi,
ok thanks for the comment. The workaround works I could also find why the histogram has zero entries: the histogram was filled with the function “AddBinContent()”. In contrast to “SetBinContent()” it seems the function is not updating the bin entries. I don’t know if this is the intended behaviour.
Philipp

[moneta]

Hi,
AddBinContent is more an internal function of the histogram, which is more efficient. If using that function you would need to set the entries by hand, calling SetEntries

Lorenzo