Ciao Patrick,
Just saw your email. Yesterday I was basically switched off from email as today untill after lunch. Let me come back to that after lunch....
-m
On 4 July 2017 at 16:46, Patrick Koppenburg patrick.koppenburg@cern.ch wrote:
Hi all,
Marcin has replied to our comments. Please see https://cds.cern.ch/record/2273007
An annoyance is that out of six assigned institutes only one has replied (me want cookie!). I wonder now how Vincenzo will decide whether this goes to EPS (or, well, I know too well).
Many replies make me unhappy.
The fits in Fig.2 are very bad at low Q. You do not mention this. How good
are your fits in terms of Q2? It shows that the parameterization is not fully correct. In particular we worry since there is a correlation between the radius and chaoticity parameters. There must be a systematic effect involved.
-> Yes, we did not mention this on perpose. We would like to avoid such a
discussion. It is known feature with this kind of parametrisation seen always in analyses from other experiments. One can hardly find a paper about BEC with such a discussion. People tried another parametrisations giving better chi2/ndf of the fit but interpretation of the parameters is then not straigthforward, and they are not quoting any results from such fits. Therefore, all the published results are quoted for Levy parametrisation or Gaussian fit allowing to compare the results. No one discussed the features of the fit in detail.
You cannot avoid such a discussion as it already started in LHCb. We see that the fit is better in the ATLAS paper (1502.07947) and in the CMS paper (1101.3518). In the latter the chi2/Ndof is given and they write "The fit quality is poor, as can be seen from the values of χ2 /Ndof . Gaussian parameterizations, which are used by some experiments, provide values of χ2 /Ndof larger than 9, which confirms the observation in [4] that an exponential parameterization is preferred." In 1005.3294 they also give fit probabilities. So not mentioning what is visible by eye is not a good strategy.
What we are worried about is that for a bad fit the value of the parameters is biased or even meaningless. The size of such a bias should be assessed.
We decided to use 2011 data only since we did not have the pythia8
simulation for 2010 sample.
This was taken 7 years ago. There was plenty of time to address that.
Cheers,
Patrick
--
Patrick Koppenburg Nikhef, Amsterdam http://www.koppenburg.ch/address.html
Bfys-physics mailing list Bfys-physics@nikhef.nl https://mailman.nikhef.nl/mailman/listinfo/bfys-physics