Choose timezone
Your profile timezone:
LHC EW precision sub-group meeting (PDFs for precision DY measurements)
→
Europe/Zurich
4/R-050 (CERN)
Description
This meeting will serve as an intermediate one between the Durham workshop and the July general meeting.
Minutes of meeting on PDFs for DY precision measurements of Tuesday 28/05/2019
=================================================================
The meeting agenda can be found at https://indico.cern.ch/event/823493/ . Brief presentations were made to summarise the conclusions from the discussions at the end of the Durham workshop and of specific follow-up exchanges addressing the two main concerns expressed in Durham, namely how to account for tolerances used in the CT and MMHT fits in the extraction of the correlations between different PDF fits from the toys and how to make sure all relevant data are available in xFitter for toy production in an appropriate way. These led to quite some discussion and to a hopefully realistic and concrete path to:
a) launch and validate toy production within the xFitter framework
b) fit the o(100) toys produced within the frameworks of the PDF fits. Concretely, these would correspond to ABM16 and NNPDF3.1 which are already public since quite some time and to private versions of CT18Z and MMHT19 which are almost ready to be published.
A number of specific points were discussed during the meeting:
1) The CT and MMHT experts present agreed that the proposal presented by Bogdan on how to account for the tolerances in the calculation of the correlations should work. This can readily be verified by comparing the spread of the toys to the official PDF uncertainties along each eigenvector. The issue of asymmetric uncertainties was also discussed and the consensus which emerged was that:
a) one would consider only symmetric uncertainties in this first stage
b) one would look at a later stage at PDF fit results displaying significant asymmetries in the uncertainties obtained (CT14 and NNPDF3.1 for eg W/Z ratios were mentioned as examples).
2) From Sasha’s presentation, it emerged that only very few datasets are currently missing in xFitter. The neutrino scattering DIS data are the most prominent one but they are already being worked on by O. Zenaiev. The other set is the electron/muon-deuteron low-energy data, but their impact at present is small, so they will only be included if considered still quite relevant by one of the PDF groups.
3) The ATLAS and CMS experimentalists who will help with the toy production and fits will organise themselves with the help of Sasha to produce and validate toy production within xFitter before exporting the toys to the PDF groups. This validation should include a quantitative comparison of the covariance matrices of the toys and of the original published data.
4) Each PDF group may find it convenient to use only a fraction of the toy data together with the nominal data for the remaining fraction of the full dataset used to produce their nominal fit to verify the input interface and the fitting of the toys step by step. CT and MMHT look forward to direct help from the experiments to do this since the experts in these groups are working with highest priority on completing the publication of their forthcoming new DPF sets.
5) As shown by Bogdan, many interesting cross-checks can eventually be performed using these toys, eg comparing the correlations between the total uncertainties in different (x.Q2) points for specific PDFs and also comparing directly actual distributions rather than their moments.
6) ATLAS and CMS (and perhaps also LHCb?) will start to take the formal steps to make sure computing resources from the experiments will be made available to produce the much larger number of fits required for a precise measurement of the correlations between the different PDFs. The requirements are particularly demanding for NNPDF where the central value of the fit of each toy can only be obtained reliably (as replica 0) from the average over ~ 100 replicas.
7) All present agreed that the results of this exercise would certainly warrant a publication which, as for other ongoing work in the precision EW working group, would be a precursor to the Yellow Report summarising the work achieved over 2018 and 2019 (and possibly somewhat beyond).
There are minutes attached to this event.
Show them.
