[Rivet] mapping of DrawNormalized() method in Root

[Andy Buckley]

Hi Debarati,

So you *want* to normalise to (unit?) area, but the Rivet normalize() 
function did not give the right result?

That should be the appropriate method, and works elsewhere. Using 
scale(histptr, 1/sumOfWeights()) will do the same thing, if every input 
event fills the histogram once.

Perhaps the issue is that you have a substantial population in the 
underflow and overflow bins, in which case the in-range normalised area 
will be significantly smaller than one. If this is the case and you 
*really* only want to normalise to the integral of the in-range bins 
(noting that this does not really give you the physical 1/sigma 
dsigma/dQ), then you can use the 3rd argument to normalize(): 
normalize(myhist, 1, false).

Does that help? If not, please send some more concrete information, e.g. 
the analysis code or YODA files.

Andy


On 08/06/16 11:27, Debarati Roy wrote:
> Hi Andy,
>              I tried Normalise() and Scale() methods in the finalise section to match with the data.
> But I could see significant mismatch.
>
> So I tried the NormalizetoSum in the plot which exhibits exact match if I do not divide the bin content by bin width which I am doing in my data histogram having variable bin width to plot normalised event per unit (1/N dN/dQ).
>
> So I am trying to look for some option so that both of these (content divided by bin width and NormalizetoSum ) can be implemented to match the data if possible.
>
> What do you think about this ?
>
> Thank you,
> Debarati
>
>
>
> ________________________________________
> From: Andy Buckley
> Sent: 08 June 2016 15:08
> To: Debarati Roy; David Grellscheid; rivet at projects.hepforge.org
> Subject: Re: [Rivet] mapping of DrawNormalized() method in Root
>
> Hi Deberati,
>
> I'm not quite sure what you want -- you want to normalise during
> plotting, or during the analysis?
>
> For the norm-in-analysis route, which we recommend, you can simply call
> normalize(myHistoPtr) in the analysis' finalize() function. You can give
> a second argument to normalise to something other than 1, or use
> scale(hist, factor) if you need to retain some of the original
> normalisation information. Note that these methods include the overflow
> bins in the normalisation: this is usually the correct thing to do, but
> is not possible if only normalising during plotting.
>
> For plotting, you can use those Normalize* attributes I mentioned
> earlier. They should be added to the relevant PLOT section in a .plot or
> .dat file, or given on the rivet-mkhtml or rivet-cmphistos command line
> like PLOT:NormalizeToIntegral=1. The full set of make-plots
> configuration directives is documented at
> https://rivet.hepforge.org/make-plots.html
>
> Andy
>
>
>
> On 08/06/16 09:51, Debarati Roy wrote:
>> Dear Experts,
>>                        Thanks for the feedback. Can you please point to where NormalizeToSum and NormalizeToIntegral are defined ?
>>
>> Thanks a lot,
>> Debarati
>>
>>
>> ________________________________________
>> From: Andy Buckley
>> Sent: 06 June 2016 17:57
>> To: David Grellscheid; Debarati Roy; rivet at projects.hepforge.org
>> Subject: Re: [Rivet] mapping of DrawNormalized() method in Root
>>
>> I agree with David, but Rivet's make-plots system does have some
>> plotting options for normalisation -- with the caveat that all it knows
>> about is the values to be rendered.
>>
>> Adding the plot key NormalizeToIntegral=1 will normalise the output plot
>> to the inferred histogram area, while NormalizeToSum=1 will scale by the
>> sum of bin values.
>>
>> But like David, I think this is much better done at the data histogram
>> level than in plotting.
>>
>> Andy
>>
>>
>> On 06/06/16 13:12, David Grellscheid wrote:
>>> Hi Debarati,
>>>
>>> I'm not sure I fully understand your question, but Rivet histograms do
>>> not know anything about how they get drawn. That's the job of a plotting
>>> tool that takes YODA files as input (rivet-mkhtml for example).
>>>
>>> A histogram can be normalized in the finalize() step of the Rivet
>>> analysis using the scale() or normalize() functions. This is done in
>>> many of the existing analyses, look there for examples.
>>>
>>> See you,
>>>
>>>      David
>>>
>>>
>>> On 05/06/2016 15:00, Debarati Roy wrote:
>>>> Dear Experts, Can you please let me know how can I implement the
>>>> DrawNormalized() method
>>>> (https://root.cern.ch/doc/master/TH1_8cxx_source.html#l02918) of Root
>>>> in Rivet for 1D histogram normalisation ? Thanks for your help in
>>>> advance.
>>>>
>>>>
>>>> Debarati
>>> _______________________________________________
>>> Rivet mailing list
>>> Rivet at projects.hepforge.org
>>> https://www.hepforge.org/lists/listinfo/rivet
>>
>>
>> --
>> Dr Andy Buckley, Lecturer / Royal Society University Research Fellow
>> Particle Physics Expt Group, University of Glasgow
>>
>
>
> --
> Dr Andy Buckley, Lecturer / Royal Society University Research Fellow
> Particle Physics Expt Group, University of Glasgow
>


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