[Rivet] Fwd: Ratio measurements in Rivet

[Andy Buckley]

Hi guys,

In case you didn't see my reply to Felix...

Yes, Gerhard is right -- at *present* Rivet can only make histograms
that come from one sort of run because the finalize() step where such
histo combination would be done is run in each run rather than as a
separate process.

The same issue means that it's hard to handle run merging for any
complex observable such as e.g. the asymmetry A-B/A+B of two
histograms... although we play a few tricks in our merging script to
identify simple normalized histos or ratios and not screw them up!

This is no different from any other analysis... and it isn't hard to
write a custom script to retrieve the numerator and denominator histos
and do the division outside Rivet. However, we are working on a
development that will allow Rivet's finalize to be run separately, which
will solve these sorts of problems for once and for all, so please make
the analyses, add all the histograms that you can, and put in a big "///
@todo <insert description>" Doxygen comment into the analysis and tell
us about it, and we will "upgrade" it when possible!

Andy


On 14/05/14 17:46, Gerhard Brandt wrote:
> 
> 
> Hi Monica,
> 
> I think nowadays responsible are Roman and Kiran, and i put the RIVET
> authors in CC for good measure.
> 
> We have indeed recently started discussing this issue with them and
> it seems that a real ratio analysis is currently not possible (at least
> it wasn't two months ago) - I will forward previous conversation.
> 
> what might be possible with some manual hacking is to run two
> consecutive RIVET analyses and then manually divide the histograms.
> I do not think this has been done, but I take it from Roman's and Andy's
> reply that this is possible...
> it does not of course correctly correlate the uncertainties
> so the usefulness is limited, basically to the central value
> 
> Gerhard
> 
> 
> On 5/14/14, 8:30 AM, Monica Dunford wrote:
>> Hi Ulla and Gerhard,
>>
>> A post-doc in my group was asking about ratios in Rivet (see below). Do
>> you know anything about the status of this in Rivet? Or know who I
>> should speak to?
>>
>> Cheers,
>> Monica
>>
>>
>>>
>>> -------- Original Message --------
>>> Subject:     Ratio measurements in Rivet
>>> Date:     Mon, 12 May 2014 12:49:35 +0200
>>> From:     Felix Müller <fmueller at cern.ch>
>>> To:     hn-atlas-Generators (General discussion and questions about
>>> ATLAS
>>> MC generators) <hn-atlas-generators at cern.ch>
>>> CC:     Shima Shimizu <Shima.Shimizu at cern.ch>,
>>> "Felix.Mueller at kip.uni-heidelberg.de"
>>> <Felix.Mueller at kip.uni-heidelberg.de>
>>>
>>>
>>>
>>> Dear Rivet experts,
>>>
>>> I'm currently implementing the Rivet routine for the inclusive jet cross
>>> section measurement at s = 2.76 TeV (ATLAS_2013_I1228693). This analysis
>>> also contains the ratio to the measurement at s = 7 TeV from 2010. The
>>> data for both, the cross section and the ratio is available on HepData
>>> (http://hepdata.cedar.ac.uk/view/ins1228693).
>>>
>>> While the implementation of the cross section at one CM energy is easy
>>> and straight forward, I'm wondering how the ratio between the two CM
>>> energies can be done. Does Rivet offer a general method to combine two
>>> datasets? Or must this combination of different CM energies happen in a
>>> post-processing step, outside Rivet?
>>>
>>> Thanks a lot!
>>> Cheers, Felix
>>>
>>>
>>
> 


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