Reimplement Atlas Z0 8 TeV lowmass

[comane]

Possible bugs were found in the old implementation:

1. The overall multiplicative factor of the luminosity uncertainty given on hep-data (see https://www.hepdata.net/record/117854) is 1.8 %. In the old implementation 1.9 % was taken

-> I have incorporated this modification as it makes not a big difference

(this branch) https://vp.nnpdf.science/XpC1u8GqSyapV3e-ffvpwg==

(master) https://vp.nnpdf.science/7KddpIE6S9u2aBAfuTybkQ==

2. In the old implementation Lumi uncertainty is assumed to be MULT UNCORR. I have noticed that setting the Lumi uncertainty to CORR the chi2 grows by an order of magnitude. I am not sure I understand this. Perhaps in this case Lumi is supposed to be UNCORR?

[Radonirinaunimi]

Hi @comane, you need to run the regenerate-data bot to make the commondata test pass.

[t7phy]

@comane besides the minor comments, seems fine to me. Pls check with @enocera (if not already done) about the CORR/UNCORR business and then this can be merged.

[enocera]

Possible bugs were found in the old implementation:

1. The overall multiplicative factor of the luminosity uncertainty given on hep-data (see https://www.hepdata.net/record/117854) is 1.8 %. In the old implementation 1.9 % was taken

-> I have incorporated this modification as it makes not a big difference

(this branch) https://vp.nnpdf.science/XpC1u8GqSyapV3e-ffvpwg==

(master) https://vp.nnpdf.science/7KddpIE6S9u2aBAfuTybkQ==

Great. This is a tiny issue.

2. In the old implementation Lumi uncertainty is assumed to be MULT UNCORR. I have noticed that setting the Lumi uncertainty to CORR the chi2 grows by an order of magnitude. I am not sure I understand this. Perhaps in this case Lumi is supposed to be UNCORR?

This is a more relevant issue. It looks like that this is a genuine bug. The luminosity must be correlated. In this very specific case, furthermore, it must be correlated with all the other ATLAS 8 TeV data obtained from the same run. Specifically, it is the 2012 run with 20.2 fb^-1. This means that the luminosity uncertainty must be tagged with a string that must be the same across the following data sets:

• ATLAS high-mass DY 2D 8 TeV [arXiv:1606.01736]
• ATLAS W ±+jet 8 TeV [arXiv:1711.03296]
• ATLAS Z pT 8 TeV (two data sets) [arXiv:1512.02192]
• ATLAS ttbar l+jets 8 TeV (all absolute diff. distr.) [arxIv:1511.04716]
• ATLAS single.inclusive jets 8 TeV [arXiv:1706.03192]
• ATLAS prompt photon 8 TeV [arXiv:1605.03495]
• ATLAS single top 8 TeV (all absolute distr.) [arXiv:1702.02859]

[t7phy]

In the ATLAS ttbar l+jets 8 TeV set, I had used MULT ATLASLUMI8. I guess the same could be used here.

[enocera]

You have freedom in choosing the string. The important thing is to use it consistently across data sets.

[scarlehoff]

Actually, I think the string we have been using is ATLASLUMI12 so it should be changed also in the top data.

[comane]

Report for last commit:
https://vp.nnpdf.science/EK2J-BJyRV2_zKT2-Zeoyg==