H1+ZEUS meeting, 18/Mar/2020
- The NLO vs NNLO comparison on slides 16&17 is not straightforward as several things different, in particular the data. Low Q^2 and trijets data are included or not in the different fits.
o It was asked if we can do a new HERAPDF2.0 NLO fit in the same way as current NNLO fit and data, to compare directly NNLO fit. This is hard as the fits are not always good and it is hard to have exactly the same data and same settings for the two orders.
o The preferred alternative is to not do a detailed comparison of NLO vs NNLO scale uncertainty in the comparison.
- The hadronisation uncertainties are worse than in the H1 paper even though more data used here.
o This is probably due to the weight given to each data set in the fit. Using the Hessian method looks better as it accounts for the weighting.
o The authors think using the Hessian could be okay but we need to choose one of three methods of correlating on p.14.
o It was noted that the hadronisation uncertainties are very different amongst the different data sets.
+ It was noted that the best way would be to run the MC simulations for all measurements and get a consistent set of hadronisation uncertainties. This is a major task and it is not obvious who can do this.
+ It was re-iterated that alpha_s is not the focus so do not want to redo all the uncertainties and we would like to choose a method which is based on what we have and is conservative.
+ A check was suggested of choosing 2% hadronisation uncertainty for all data points.
~ It was discussed that at low Q^2 it is larger than 2%, say 4-5%.
o As a cross check, the authors will choose some reasonable values, e.g. 4-5% at low Q^2 and 2% at high Q^2 and see how this affects the results.
o The question of offset or Hessian method was discussed. However, if Hessian, then use the correlated approach.
o Authors to send around slides to EB with study and proposal of which method to choose. We will try and decide by e-mail and come to an agreement of which method to use. If agreement cannot be met like this and discussion needed, we will set up a meeting to address this issue.
- The plot of the decrease in gluon uncertainty on page 8 was questioned, since no decrease had been seen before (p.77 of https://indico.desy.de/indico/event/11010/contribution/1/material/slides/0.pdf) but it was established that this plot is for total uncertainties not experimental. For experimental uncertainties, there is essentially no difference. The change in total uncertainties derives from a change in procedure for model uncertainties, which avoids a certain amount of double counting of Q20 and m_c variations. In any case the change is not dramatic. No action needed.
- The difference between in alpha_s of 0.1151 and 0.1158 after adding the extra 6 points of low pt, high Q2 H1 data was questioned. It was asked if these points pull against the rest - this will be estimated, for information only. It was also suggested that maybe differing estimates of hadronisation uncertainty contribute to this.