[Rivet] Weights in Rivet

[Andy Buckley]

On 20/10/14 15:03, Stefano Frixione wrote:
> Hi Andy,
> so I gather that the short answer for *existing* Rivet analysis
> is "no"?

Correct! I'm not good at the short answer ;-)

The way that we will add it will be to the underlying framework so that
the analysis codes will only need minimal changes (specifically they
will not deal in event weights anymore: that will be entirely handled by
the Rivet system)

> Concerning:
>> It would be great if you could let us know how you would like them to be
>> used by default: our plan at the moment is that we write out N copies of
>> every histogram as a weight-variation, and build some logic into our
>> default plotting tools to turn these groups of variations into displayed
>> error bands. This requires that we can clearly identify the groups and
>> tell how to combine them... which is specified at the LHE level, but not
>> the HepMC one -- right? So the first version might be a bit approximate,
>> e.g. always a min-max envelope rather than sometimes a 68% interval,
>> sometimes a quadrature sum, etc.
> in my opinion, plotting directly the error bands is error prone and
> not particularly flexible. It is fine, but one should *also* get
> a copy of all desired histograms for each of the scale and PDF
> choices, so that a-posteriori manipulations could be done by the user.
> For example, you may want a scale error band that corresponds to
> all N*N choices, another one thar results only from setting muF=muR,
> and so forth. Likewise, one may want to experiment with PDF
> combinations. Needless to say, I'm open to suggestions.

Ok, that's good feedback -- and makes sense of course. We distinguish
between histogrammed data, and the summary info required for plotting
it, so this separation should automatically be the case; plotting bands
by default is only a presentational issue and the individual lines
within the band will be available in the analysis output .yoda file. I'm
sure we'll have to iterate a bit to get the best default behaviour.

Andy

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