[Rivet] CMS RIVET analyses

[Andy Buckley]

On 18/11/13 15:59, Nhan V Tran wrote:
> Hi Andy,
> 
> Thanks for following up on the CMS jet mass RIVET routine last week.  I forgot to ask you, one of the other CMS conveners was asking me...
> 
> Can you let me know what CMS analyses are in the RIVET queue and what's the expected time to be included?

Hi Nhan, (and CC'ing the Rivet dev list)

I finally got time today to really spend some time on incorporating the
various new analyses for Rivet 1.8.4. We'll release this very soon,
followed as soon as we can by an equivalent 2.0.1.

There are two CMS analyses in our queue: your jet mass one, and
CMS_2013_I1258128.cc (rapidity distributions in Z+jet and gamma+jet
events). If there are more that we can include right away, now is the
time to let us know!

I have a few questions/comments about your analysis, actually: we don't
just copy them into place but do a bit of code checking and clean-up
first... which is why it's usually not a 100% straightforward operation.
A bit annoying, I know, but once we've got them in the system we have to
maintain them... hence we sometimes ask for fixes and clarifications!
Ok, here goes:

* I've split the analysis into three separate analyses, with _ZJET,
_WJET, and _DIJET suffixes. The _ZJET one is the one that I've tidied up
so far (e.g. putting the pT bin searching into a function really
compresses the code blocks for the three jet types), and I've attached
it for your interest. (It quite possibly doesn't compile right now: it's
the code logic that I'd like feedback on, if you have any.)

* 2D histos will coming soon in Rivet v2 / YODA. They are an order of
magnitude more fiddly than 1D ones, if you really want to handle
overflows and so-on nicely. I though that they worked already in Rivet
v1, though... but HepData doesn't (yet) support them so I think this 1D
slice approach is better anyway.

* There is a bit of code to treat any negative weights as equivalent to
1.0. Why is that? I expect it's a bad idea, especially with more NLO
simulations becoming normal, but I wanted to check rather than just
remove it.

* The back-to-back requirement between the W/Z and the leading jet i
actually applied in reverse:

  // Get the leading jet and make sure it's back-to-back with the Z
  const fastjet::PseudoJet& j0 = psjetsCA8_zj[0];
  if (!isBackToBack_zj(zfinder, j0)) { //< @todo ARGH!?!

(This is my tidied version of the code, but I checked that this isn't a
problem that I introduced!) Is this a (repeated) typo, or is the
back-to-back cut *really* meant to work this way round? (Which would be
strange.) The isBackToBack function contents do match its name.

* Lastly, something on which we'll maybe need input from other Rivet
developers: you're using explicitly FastJet3 functionality to groom the
jets. That's great: you've written the first official experiment Rivet
analysis to require FJ3! But it means that we will need to require FJ3
for any Rivet version which includes this analysis... which we currently
don't do. I think we can happily do that for Rivet version 2, but
perhaps it's not the best idea to make that change in our last ever 1.x
minor version.

I suggest that we release Rivet 1.x with the existing FJ > 2.x
requirement, and also make your analysis code public on the Rivet
website for FJ3 users to build. Or we could make a special exception and
build the analysis only if FJ2 is present: that might be the neatest
option. For Rivet 2.0.1 we can make FJ3 a build requirement: I think it
is time for that anyway, but thanks for giving us a concrete reason to
do so!

Cheers, and please let me know about those queries in the code and if
you have any other feedback on the "tidied" version: once I know what to
do I'll carry on doing similar tidying for the _DIJET and _WJET source
files.

Andy

-- 
Dr Andy Buckley, Royal Society University Research Fellow
Particle Physics Expt Group, University of Glasgow / PH Dept, CERN
-------------- next part --------------
A non-text attachment was scrubbed...
Name: CMS_2013_I1224539_ZJET.cc
Type: text/x-c++src
Size: 7998 bytes
Desc: not available
URL: <http://www.hepforge.org/lists-archive/rivet/attachments/20131122/092e3712/attachment.cc>