[Rivet] rivet: normalizing histograms to variable bin width

[Andy Buckley]

On 17/03/14 14:09, Oldrich Kepka wrote:
> Hi Andy,
> 
> 
> On 17 March 2014 13:20, Andy Buckley <andy.buckley at cern.ch> wrote:
>> Hi Oldrich (and CC'ing the Rivet list -- please send to that rather than
>> me directly!),
>>
>> In the API of YODA the bin width is taken into account when calling the
>> HistoBin::height() method:
>>
>> https://yoda.hepforge.org/trac/browser/include/YODA/HistoBin1D.h#L97
>>
>> (and a corresponding divide by x-y bin area in HistoBin2D)
>>
>> By "normalize to bin width", I'm guessing that you mean "divide by
>> integral = sum of bin & overflows weight sums". For bins, the sum of
>> weights is the area rather than the height. Is that what you had in
>> mind? If so (and I'm sure this is the physically correct approach) then
>> this is already what YODA does:
>>
>> Histo1D::normalize:
>>   https://yoda.hepforge.org/trac/browser/include/YODA/Histo1D.h#L146
>> calls Histo1D::integral:
>>   https://yoda.hepforge.org/trac/browser/include/YODA/Histo1D.h#L298
>> calls Histo1D::sumW:
>>   https://yoda.hepforge.org/trac/browser/src/Histo1D.cc#L43
>>
>> Does that answer your question?
> 
> 
> In the mkScatter function that converts Histo1D etc. to
> TGraphAsymmErrs-like Scatter2Ds.
> 
> This answers my question. I knew I must have been missing something
> fundamental. Histograms are converted to scatter, where the height()
> comes into play, so that the plotted result is a differential cross
> section.

Exactly. It's "just cosmetics" that when we plot we account for the bin
width measure... but it's the *right* cosmetics ;-)

Andy


> we use the
>> Bin::height() function so that the plotting will look correct, but
>> internally the area() == sumW() are much more important.
>>
>> Andy
>>
>>
>> On 17/03/14 12:38, Oldrich Kepka wrote:
>>> Hi Andy,
>>>
>>> sorry for this basic question. Is there already an implementation of a
>>> method normalizing histograms to produce differential cross section,
>>> i.e. normalizing to bin width.
>>>
>>> Presumably, there must be, because there are several differential
>>> cross sections implemented. I am probably missing something, but I
>>> cannot find the place where the bin widtf is taken into account
>>> (looking through Histo1D, Bin1D, Dpn1D ... ). The usual Analysis
>>> scale/normalize methods boil down to scaleW of Dbn1D; scaling
>>> according to global desired factor.
>>>
>>> Otherwise, I would go for something like this.
>>>
>>>      _h_ZZ_ZpT.scale( norm);
>>>       for(size_t i = 0; i< _h_ZZ_ZpT-> numBins(); i++) {
>>>                 _h_ZZ_ZpT->bin(i). scaleW( 1/_h_ZZ_ZpT->bin(i). width() );
>>>          }
>>>        normalize(_h_ZZ_ZpT    , 1);
>>>
>>> http://hepdata.cedar.ac.uk/view/ins1203852
>>>
>>>
>>> Thanks,
>>> Oldrich
>>>
>>
>>
>> --
>> Dr Andy Buckley, Royal Society University Research Fellow
>> Particle Physics Expt Group, University of Glasgow / PH Dept, CERN


-- 
Dr Andy Buckley, Royal Society University Research Fellow
Particle Physics Expt Group, University of Glasgow / PH Dept, CERN