[Rivet] D0 2001 W/Z analysis

[Andy Buckley]

Lars Sonnenschein wrote:
> Hello all
> I have put high statistics plots with HerwigJimmy & ATLAS + W/Z Tune
> as well as with Pythia 6413 with Tune A
> at
> http://www-d0.fnal.gov/~sonne/D0_2001_S4674421/
> D0_2001_WptZpt_HerwigJimmy6510_ATLAS_and_WZtune_datnormed_RootRatio_partons.eps 
> and 
> D0_2001_WptZpt_Pythia6413_TuneA_datnormed_RootRatio_partons.eps       
> 
> The simujation is normed to data in the first three plots and the ratio I 
> have still computed with ROOT. 

I'd say those look like pretty good agreements both to the data and each
other.

> Somehow I fail to manage to get the ratio written out correctly.
> 
> If I put the ratio into a datapointset it is written out but all data 
> points and errors are zero.

How are you putting the ratio into a DPS? The Rivet code that I wrote to
"normalise" the AIDA tree to DPS objects works fine (since we get
histograms at all!)

> If I write out a profile histogram, which I fill after the division there 
> are problems due to the errors (their should be at least two entries per 
> bin to get a mean and a variance.)

Why would you want to use a profile histogram? It's completely wrong for
this!

> Modifying the bins directly is forbidden (l-value).

That's what DPSes are for.

> Adding a special function to the LWH histo class does not work either.
> It seems not be used but the AIDA class directly, therfore already the 
> normalization and integral functions of the LWH class cannot be used.

I don't understand this... of course the AIDA classes don't use a
function that hasn't been defined until now. You would have to use that
function in the Rivet code that performs the normalisation and
conversion to DPS. But I prefer the approach which uses a DPS, since
that's what the division should be producing anyway.

> Either I simply remove the ratio histo from the analysis and everybody can 
> do the divison on its own after running Rivet, or I would need to touch 
> AIDA or I would need to touch the division.

The ratio plot should be in the analysis --- it has the virtue of being
an experimental measurement that avoids the LO/NLO cross-section
normalisation that makes the other plots need explicit normalisation to
the data and is useful for tuning and validation. It should definitely
be there in the Rivet output.

The approach with having a DPS for the ratio in the analysis code is the
nicer one to follow and I'm not aware of any technical reason why it
wouldn't work, so I suggest that you investigate that further. And if
there is a problem, we should find it and fix it, rather than ignore the
issue!

Andy

-- 
Dr Andy Buckley
Institute for Particle Physics Phenomenology
Durham University
0191 3343798 | 0191 3732613 | www.insectnation.org