[Rivet] question to yodamerge

[Frank Siegert]

Hi Hannes,

Ah, of course... the normalize also uses a scale in the end, so in
principle the yodamerge should even work for normalized histos. I
don't quite understand why there is no ScaledBy annotation in that
case. Can you put the yoda files that come out of your individual runs
somewhere for us to look at?

Cheers,
Frank



On 14 April 2015 at 09:44, Hannes Jung <hannes.jung at desy.de> wrote:
> Hi Frank
>
> thanks a lot for your reply.
>
> I fully agree with your Note 1 .... it took me quite a bit
> to find out what to do to get the merging....
>
> The funny thing with this particular histo is, that yodamerge
> works for files generated with one process, but gives the
> warning for another process.... can it be, that there is something
> with the number of entries in the histo or soemthing similar ?
>
> Cheers
> Hannes
>
>
>> On 14.04.2015, at 09:35, Frank Siegert <Frank.Siegert at cern.ch> wrote:
>>
>> Hi Hannes,
>>
>> this particular histogram seems to be normalised to a given integral
>> (which I don't claim to understand:
>> 0.7*crossSection()*_sumW/sumOfWeights()) instead of scaled
>> proportional to the cross section like dsigma/dX. This means there is
>> a priori no clear prescription of how they can be merged from
>> different runs, and thus the simple yodamerge script fails. If there
>> is some way to merge these objects even though they are normalised
>> then you have to write a small Python script yourself which does it.
>>
>> Note 1: The --assume-normalized switch has an unfortunate naming and
>> has been a constant source of confusion for users. This has been
>> discussed before, and I'll state my opinion again:
>> - the switch should be the default behaviour of yodamerge (because
>> most of our distributions are of the dsigma/dX type which need it)
>> - there should be two options to overwrite (/replace) the broken
>> heuristic of whether something should be merged or added:
>> --force-merge (if you have multiple identical runs from different
>> random seeds) vs. --force-add (if you have multiple samples that have
>> to be added, e.g. electron and muon channel, Np0...Np4, ...).
>>
>> Note 2: On the proper and automatic run combination for arbitrary
>> objects (including arbitrarily normalised histograms) we have this
>> ticket:
>> https://rivet.hepforge.org/trac/ticket/457
>>
>> Cheers,
>> Frank
>>
>>
>> On 14 April 2015 at 06:57, Hannes Jung <hannes.jung at desy.de> wrote:
>>> Dear Rivet experts
>>>
>>> i use yodamerge to add yoda files from stat independent runs
>>> with
>>>
>>> yodamerge --assume-normalized  .....
>>>
>>> now in some cases I get the warning:
>>>
>>> WARNING: Abandoning normalized merge of path /CMS_2013_I1256943/d01-x01-y01 because not all inputs have ScaledBy attributes
>>>
>>> and the results are weird.
>>>
>>> What is wrong there, or what could I check ?
>>> Thanks a lot
>>> cheers
>>> Hannes
>>>
>>> ***********************************************************************
>>> Hannes Jung
>>> Email: Hannes.Jung at desy.de
>>> mobile :+49 40 8998 93741
>>> http://www.desy.de/~jung
>>> Tel: +49 (0) 40 8998 3741
>>> Fax: +49 (0) 40 8994 3741
>>> DESY, CMS 01B/02.213
>>> Notkestr.85, 22603 Hamburg, FRG
>>> ***********************************************************************
>>>
>>>
>>>
>>> _______________________________________________
>>> Rivet mailing list
>>> Rivet at projects.hepforge.org
>>> https://www.hepforge.org/lists/listinfo/rivet
>
> ***********************************************************************
> Hannes Jung
> Email: Hannes.Jung at desy.de
> mobile :+49 40 8998 93741
> http://www.desy.de/~jung
> Tel: +49 (0) 40 8998 3741
> Fax: +49 (0) 40 8994 3741
> DESY, CMS 01B/02.213
> Notkestr.85, 22603 Hamburg, FRG
> ***********************************************************************
>
>
>