[Rivet] Rivet Jet Analysis

[Andy Buckley]

On 26/01/11 14:27, ST Harnew wrote:
> Dear Andy,
> 
> I am a student currently working on my final year MSci project at
> Bristol University. The aim is to find tunes for Pythia8 using the Rivet
> and Professor packages.  I understand from my supervisor Nick Brook that
> you have had a large contribution in the making of Rivet, and I have
> noticed that you made a considerable number of the analysis.  I have
> recently run into difficulties with some of these, and I would be
> extremely grateful if you could point me in the right direction.

No problem! I'm copying this to the Rivet mailing list, as that way
you'll get a collective response from various Rivet developers and
contributors.

> I have just starting looking at jet analysis, in particular
> CDF_2001_S4751469, CDF_2004_S5839831, CDF_2008_LEADINGJETS. All these
> analyses request that they be run with several pT cutoffs to fill the
> histograms. When I tried running some of these, Rivet fails while
> creating the hisograms...
>
> Event 999600 (7989 s elapsed)
> Event 999700 (7989 s elapsed)
> Event 999800 (7990 s elapsed)
> Event 999900 (7991 s elapsed)
> Event 1000000 (7992 s elapsed)
> Finished event loop
> python: ../../include/LWH/Profile1D.h:362: virtual double
> LWH::Profile1D::binError(int) const: Assertion `denom > 0' failed.
> Abort
> 
> 
> This is a problem for some of the pT cutoff runs, and most of the
> MinBias runs.

What version of Rivet are you using? I recall that we had a release in
which there was a problem of this kind in the histogramming, but I'm
pretty sure this was fixed in a more recent release. If you're not using
version 1.4.0, please upgrade to it!

> For the minimum bias runs, I am using the standard softQCD:minbias
> setting, and for the pT cuts I am using hardQCD:all along with the
> appropriate cut-off, PhaseSpace:pTHatMin, as listed in the analysis
> descriptions.
> 
> For the CDF_2008_LEADINGJETS  <Nch> against pT(leading jet) plots, the
> runs from each pT cut are needed to fill a single histogram. Should I
> deal with this by chopping each histogram down to its valid region (say
> 0-20 pT for the minimum bias run, then 20-100pT for the pTmin=10GeV run
> etc) then stick all these together to make a complete distribution? Or
> is there a smarter way of doing this.

Heh... there is a smarter way, but what we currently do for most studies
is similar to what you said! The smarter way is to use weighted events,
where the probablility of generating an event at high-pT is boosted by a
factor of (say) pT^5, and the event weights are down-scaled by the same
factor so that the Rivet observables are still physically sensible.

This way you can probably cover the whole pTlead spectrum in that
analysis with a single run (and possibly lower statistics, although
running over a million or so events is ~always a good idea IMO).

You might need to do something special for the min-bias region of the
plots, as I know that in PYTHIA 6 switching to a pTmin of 0 will disable
that event weighting... I'm not sure what the situation is for Pythia 8,
but there are people on this list who should certainly know :)

Cheers, and good luck!
Andy

PS. We'll also have a new version of Professor (1.2.0) out very soon...
in the next day or so if we can, depending on whether or not I get some
requested bug reports. It'll be very much new and improved -- please
keep an eye on the website and use it when it appears!

-- 
Dr Andy Buckley
SUPA Advanced Research Fellow
Particle Physics Experiment Group, University of Edinburgh

The University of Edinburgh is a charitable body, registered in
Scotland, with registration number SC005336.