TTB integrated x-sec datasets

[t7phy]

No description provided.

[scarlehoff]

Thanks for this! I've done a first test, I find some differences in the data. When this difference is expected please write something about it in the PR (in order to know whether it is a bug or a feature).

For Atlas only the 7 TeV one agrees for everything. I find some differences in the t0 chi2 which is pointing to a difference in the treatment or MULT and ADD?

 > Testing old: ATLASTTBARTOT7TEV vs new: ATLAS_TTBAR_7TEV_TOT_X-SEC
The t0 chi2 is different: 2194.2430166777112 vs 1102.1581162540726

 > Testing old: ATLASTTBARTOT8TEV vs new: ATLAS_TTBAR_8TEV_TOT_X-SEC
Data differences under 1e-2, continuing comparison...
	 242.4  242.9
# Differences in the computation of chi2 391.76 vs 548.71
    The covmats are different
    even the diagonal
    
 > Testing old: ATLASTTBARTOT13TEV vs new: ATLAS_TTBAR_13TEV_TOT_X-SEC
Relative differences above 1e-2                 
	 818.0   830 

For CMS instead I find perfect agreement for everything but the 13 TeV one, where there's a big difference in the results:

792.0   746 

So for now I've done the following report only for CMS: https://vp.nnpdf.science/WEzTkruoTt6kZNpn8igKdA==

[t7phy]

Thanks for this! I've done a first test, I find some differences in the data. When this difference is expected please write something about it in the PR (in order to know whether it is a bug or a feature).

For Atlas only the 7 TeV one agrees for everything. I find some differences in the t0 chi2 which is pointing to a difference in the treatment or MULT and ADD?

 > Testing old: ATLASTTBARTOT7TEV vs new: ATLAS_TTBAR_7TEV_TOT_X-SEC
The t0 chi2 is different: 2194.2430166777112 vs 1102.1581162540726

 > Testing old: ATLASTTBARTOT8TEV vs new: ATLAS_TTBAR_8TEV_TOT_X-SEC
Data differences under 1e-2, continuing comparison...
	 242.4  242.9
# Differences in the computation of chi2 391.76 vs 548.71
    The covmats are different
    even the diagonal
    
 > Testing old: ATLASTTBARTOT13TEV vs new: ATLAS_TTBAR_13TEV_TOT_X-SEC
Relative differences above 1e-2                 
	 818.0   830 

For CMS instead I find perfect agreement for everything but the 13 TeV one, where there's a big difference in the results:

792.0   746 

So for now I've done the following report only for CMS: https://vp.nnpdf.science/WEzTkruoTt6kZNpn8igKdA==

Actually I am pretty confident about the data and uncer. here because there is no data implementation here, everything is taken just from arxiv papers directly and I jut rechecked the values, it does appear to be consistent. Why does the old implementation differ? That I am not sure tbh. A possibility is that values have changed on arXiv due to improved lumi...??

[scarlehoff]

There's a comment about a fix in the old implementation https://github.com/NNPDF/nnpdf/blob/master/nnpdfcpp/data/commondata/DATA_CMSTTBARTOT13TEV.dat

Maybe @enocera remembers

[felixhekhorn]

A (slightly) unrelated comment (but since I saw this in my newsfeed I decided to post it nevertheless - feel free to discard it as off-topic): I assume this is about fully inclusive ttbar xs (and the jets are in this PR by mistake), right?

I'm currently working on MaunaKea a.k.a. "Hawaiian top++" patched on top of top++ by Czakon and Mitov for my work here in Jyväskylä - do we want to use this eventually on the theory side? The theory is currently provided by mg5 which means it is limited to NLO, but I could go full NNLO. The impact on the fit will be 0 since we talking about the grand total of 7 datapoints (associated to this PR).

[scarlehoff]

Yes, I think that will be great! I mean, just for consistency sake it is worth it.

[t7phy]

A (slightly) unrelated comment (but since I saw this in my newsfeed I decided to post it nevertheless - feel free to discard it as off-topic): I assume this is about fully inclusive ttbar xs (and the jets are in this PR by mistake), right?

The jets appear in the commits because the branch that is collecting the datasets already has the jets merged as they were considered finalized and done

[scarlehoff]

To be corrected in the list of legacy dataset names:

ATLAS_TTBAR_13TEV_TOT_X-SEC should be compared against ATLAS_TTBARTOT_13TEV_FULLLUMI

When doing this change, everything is ok.

The difference for the CMSTTBARTOT13TEV vs CMS_TTBAR_13TEV_TOT_X-SEC remains though and should be understood. The difference is 792.0 746
We need to understand whether maybe CMS_TTBAR_13TEV_TOT_X-SEC and CMSTTBARTOT13TEV actually correspond to different datasets.

[t7phy]

@scarlehoff

The difference for the CMSTTBARTOT13TEV vs CMS_TTBAR_13TEV_TOT_X-SEC remains though and should be understood. The difference is 792.0 746 We need to understand whether maybe CMS_TTBAR_13TEV_TOT_X-SEC and CMSTTBARTOT13TEV actually correspond to different datasets.

So nnpdf uses https://arxiv.org/abs/1510.05302, which is exactly what I have used too:
the values on arxiv are:
v1: 769
v2 and v3: 746
I am unable to find 792 anywhere. Is 'CMSTTBARTOT13TEV' used in the 4.0 runcard?

[scarlehoff]

Yes

nnpdf/n3fit/runcards/reproduce_nnpdf40/NNPDF40_nnlo_as_01180_1000.yml

Line 74 in 72008b2

- {dataset: CMSTTBARTOT13TEV, frac: 0.75, cfac: [QCD]}

https://github.com/NNPDF/nnpdf/blob/master/nnpdfcpp/data/commondata/DATA_CMSTTBARTOT13TEV.dat