[Rivet] Question about YODA Profile1D: how to show RMS instead of errors of the mean

[Andy Buckley]

On 13/06/15 11:37, Adrian Buzatu wrote:
> Hi RIVET/YODA authors,
> 
> For the studies I am doing on the jet energy scale it was suggested to
> me to reproduce some YODA Profile1D where instead of the error bars on
> the mean to show the RMS of the distribution, as we already have the
> mean on the Y axis. Do you know how to do that?
> 
> On a different, but maybe related note, do you know if I can interface
> ROOT with RIVET to use ROOT insted of YODA? I am familiar with how to do
> this in ROOT, and then use PyROOT, maybe that can help. But I hope plan
> A with YODA has a feature for this.

Hi Adrian,

There's some confusion here about what YODA does: it mostly just handles
the statistical definition of histograms, but the plotting is external.

rivet-cmphistos converts to the simple make-histos format, and the
decision about what size to give to error bars is built into the
converter code. It doesn't have a simple flag to change the profile
error bar plotting, though.

Our future direction for Rivet and YODA plotting is to use the Python
matplotlib (mpl) library. There's the start of some helper functions in
the yoda.plotting Python module, but even without touching those it's
pretty easy to do mpl plotting of YODA objects:

  ## Load a profile histo
  import yoda
  aos = yoda.read("/home/andy/proj/hep/rivet/Rivet.yoda")
  p = aos["/MC_GENERIC/EtaSumEt"]
  assert type(p) is yoda.Profile1D

  ## Get data points and errors
  xs, ys, exs, eys, sigys = [], [], [], [], []
  for b in p.bins:
      if b.sumW:
          xs.append(b.xMid)
          exs.append(b.xWidth/2)
          ys.append(b.mean)
          eys.append(b.stdErr)
          sigys.append(b.stdDev)

  ## Plot
  import pylab as pl
  pl.hold(True)
  pl.errorbar(xs, ys, xerr=exs, yerr=sigys, marker="None", ls="None",
ecolor="black")
  pl.errorbar(xs, ys, xerr=exs, yerr=eys, marker="None", ls="None",
ecolor="red")
  pl.plot(xs, ys, "o", color="black")
  pl.savefig("rmsprofile.pdf")

With the demo Rivet.yoda file that I had lying around, this produced the
attached PDF (note that I used the standard deviation rather than RMS
for this plot, since b.rms() wouldn't subtract the mean value in the
calculation of the distribution width). It should be pretty easy to
extend that script, and I can help you to use the yoda.plotting tools if
you like. I'm also going to add an optional argument to the Profile1D
mkScatter, function, which will make it possible to do this sort of
plotting in just a few lines.

There's also a yoda2root script, which makes use of converter functions
in the YODA/ROOTCnv.h header... I'm not sure if it works completely for
profiles, though, since ROOT's interface for populating those is very
odd. If you go that route, let us know how you get on!

Andy

-- 
Dr Andy Buckley, Lecturer / Royal Society University Research Fellow
Particle Physics Expt Group, University of Glasgow
-------------- next part --------------
A non-text attachment was scrubbed...
Name: plotprofile.py
Type: text/x-python
Size: 703 bytes
Desc: not available
URL: <https://www.hepforge.org/lists-archive/rivet/attachments/20150613/1e76166e/attachment.py>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: rmsprofile.pdf
Type: application/pdf
Size: 33307 bytes
Desc: not available
URL: <https://www.hepforge.org/lists-archive/rivet/attachments/20150613/1e76166e/attachment.pdf>