Theory 400 needs to be recomputed or rechecked.

[scarlehoff]

I have recomputed all the fktables for theory 400 by using the grids that @andreab1997 has in dom for theory 424.

The only fktables that I could not recompute satisfactorily are:

• JETS (because of the known reasons)

• CMS_SINGLETOP_TCH_R_8TEV

• ATLAS_SINGLETOP_TCH_R_7TEV

• CMS_SINGLETOP_TCH_R_13TEV

• ATLAS_SINGLETOP_TCH_R_13TEV

The grids for these datasets do exist in Theory 424, however they don't correspond to the fktables in theory 400 (so the grids for those fktables came from somwhere else)

---

What I find more worrying though is that I found a number of differences when recomputing the tables. These differences could be on my side but they need to be understood since I will need to recompute the baseline fit for theory 400.

With a threshold of relative difference of 1e-1 (which is quite high) I find differences in the following datasets: ['BCDMSP_dwsh', 'BCDMSD_dw_ite', 'CHORUSNBPb_dw_ite', 'NTVNUDMNFe_dw_ite', 'NTVNBDMNFe_dw_ite', 'HERACOMBNCEP460', 'HERACOMBNCEP575', 'HERACOMBNCEP820', 'HERACOMBNCEP920', 'HERACOMB_SIGMARED_B', 'ATLAS_TTB_DIFF_8TEV_LJ_TRAPNORM', 'ATLAS_TTB_DIFF_8TEV_LJ_TTRAPNORM', 'CMS_TTBAR_2D_DIFF_MTT_TRAP_NORM']

These difference are as high as 20% for some datapoints. I'm using cuts fromfit so all the points for which I find differences are included in the fit.

Some of the differences, such as CMS_TTBAR_2D_DIFF_MTT_TRAP_NORM I understand. The grid in theory 424 is simply outdated I can recompute. Some others, however, are more difficult to reconcile.

For instance, for HERACOMB_SIGMARED_C we see:

         th_400        new           ratio
15    0.307621  0.252231  1.219599                                                        
16    0.273128  0.227444  1.200859                                                        
17    0.235284  0.200045  1.176152                                                        
18    0.198524  0.172997  1.147562                                                        
19    0.153816  0.139164  1.105284                                                        
20    0.111205  0.105673  1.052355                                                        
21    0.323548  0.287088  1.126997                                                        
22    0.287826  0.258962  1.111459                                                        
23    0.244099  0.224003  1.089715                                                        
24    0.200224  0.188170  1.064059                                                        
25    0.154074  0.149387  1.031377                                                        
26    0.115681  0.115838  0.998646                                                        
27    0.356254  0.356238  1.000046                                                        
28    0.324555  0.324540  1.000046                                                        
29    0.270971  0.270959  1.000041                                                        
30    0.221904  0.221894  1.000046                                                        
31    0.197241  0.197232  1.000044                                                        
32    0.154316  0.154310  1.000041                                                        
33    0.127509  0.127504  1.000040                                                        
34    0.340235  0.340203  1.000093                                                        
35    0.295819  0.295791  1.000095                                                        
36    0.244806  0.244783  1.000093                                                        
37    0.199985  0.199966  1.000090                                                        
38    0.156639  0.156625  1.000088                                                        
39    0.106758  0.106749  1.000082                                                        
40    0.357800  0.357756  1.000125                                                        
41    0.296627  0.296590  1.000124                                                        
42    0.227946  0.227918  1.000120                                                        
43    0.163668  0.163649  1.000115                                                        
44    0.085586  0.085577  1.000109                                                        
45    0.266621  0.266585  1.000134                                                        
46    0.152677  0.152658  1.000125                                                        
47    0.198313  0.198285  1.000138                                                        
48    0.102014  0.102001  1.000130                                                        
49    0.182099  0.182073  1.000142                                                        
50    0.088431  0.088419  1.000137                                                        
51    0.062263  0.062255  1.000122    

The first group of points is very different while the others agree perfectly. Which fktable is correct? The one in theory 400 or the new one?

[andreab1997]

@scarlehoff Can you add the Q2 values to that table please? Because if, for the points that do not agree, is under ~2.75, we just do not care.

Anyway, since we are using the same version of eko, I believe there are only two possible sources for this difference:

1. Different pineappl version, which means basically the new evolve. However I believe that @cschwan carefully tested the new function so I believe most likely the problem is not there.
2. Different grids. This can well be the case because we do not have an official source of grids and yaml files. In this case I believe that my grids are newer (just because theory 400 has been computed before) and so most likely "more correct".

Las thing that maybe you can easily check is if your fks predictions match the grid predictions (you can do it using pineko compare). If not than most likely the problem is 1 (but I am not expecting this to be true).

[scarlehoff]

@andreab1997 There are many points that do not agree in many tables. These points are included in the fit so we do care. In the HERA that I've put in OP Q starts at 12 for instance (for point 15) HERA_NC_318GEV_EAVG_SIGMARED_CHARM.pineappl.lz4

Different pineappl version, which means basically the new evolve. However I believe that @cschwan carefully tested the new function so I believe most likely the problem is not there.

I've only used evolve for one of the datasets. The rest is all using pineko normally.

Different grids. This can well be the case because we do not have an official source of grids and yaml files. In this case I believe that my grids are newer (just because theory 400 has been computed before) and so most likely "more correct".

If this is the case, we need to understand whether this is the case and then recompute the whole of theory 400 and make sure that we are all using the same grids. Otherwise we are wasting a lot of time...

if your fks predictions match the grid predictions

They do. Also for the fktables for which I've used evolve.

[scarlehoff]

I've added everyone else to this issue. I believe Theory 400 needs to be probably recomputed and surely rechecked. Luckily is not the entire set of datasets.

[andreab1997]

I am still not sure but it seems that I found something. I have recomputed the grid for HERACOMB_SIGMARED_C and checking its lumis with pineappl it seems that it only contains contributions up to 4 (so the charm). However, the grid that I am using contains also the bottom contribution and this could explain the difference that @scarlehoff found. Moreover, all the nf related parameter in the theory card were setted to 5 but at the end the grid lumi was only up to 4 so it seems that yadism ignore those parameters and just computes with 4. @AleCandido @felixhekhorn @cschwan Do you have any idea?

I tried to compute the fktable starting with the nf=4 grid. If it matches the theory 400 fk, it would mean that in theory 400 we used to use nf=4 grids but then for some reason we produced new grids with nf=5. I don't know in which way but I believe this is the cause of the issue.

[cschwan]

When you say 'grid' do you mean FK tables or the grids at process scale?

[andreab1997]

When you say 'grid' do you mean FK tables or the grids at process scale?

I mean grid

[cschwan]

I just noticed that https://github.com/NNPDF/pineapplgrids doesn't have the DIS grids, and there's no pull request for them; I believe this is where we should store all (unevolved) grids, together with a pull request in which the predictions are compared against the old results. Can someone do that? I can help with it. Every prediction that isn't available in this form should be considered unvalidated and dangerous.

[scarlehoff]

I will get in one folder all the correct grids and the corresponding fktables. Once I know what the correct grids are, which is yet not the case. Also, not sure what to compare.

[cschwan]

I will get in one folder all the correct grids and the corresponding fktables. Once I know what the correct grids are, which is yet not the case.

Great!

Also, not sure what to compare.

I believe we can compare the old DIS FK tables (which I believe is how they were stored) with the new grids. I can make the comparisons.

[scarlehoff]

I believe we can compare the old DIS FK tables (which I believe is how they were stored) with the new grids. I can make the comparisons.

But they are not supposed to be equal. That's one of the problems.

[felixhekhorn]

I just noticed that https://github.com/NNPDF/pineapplgrids doesn't have the DIS grids, and there's no pull request for them; I believe this is where we should store all (unevolved) grids, together with a pull request in which the predictions are compared against the old results. Can someone do that? I can help with it. Every prediction that isn't available in this form should be considered unvalidated and dangerous.

The structure there is not sufficient for DIS, since NNLO (FONLL-C) ones do not contain the NLO (FONLL-B) ones - that being said I believe we should split by theory and link whenever possible (as we do in pineko)

(the "final place" might well still be a git repo)

[scarlehoff]

I will create a theory_400 folder with all the grids so at least Theory 400 can be reproduced.

That said, before thinking on where to store it I would like to know what to store...

[cschwan]

We can make a comparison at NNLO. In fact we should make the comparisons at the highest available order. In general I believe that even if there are differences it's important to document them, and more importantly we need a source of validated theory predictions. Otherwise this issue will happen again.

[cschwan]

@scarlehoff same holds true for the fixed-target Drell-Yan grids.

[scarlehoff]

I will add everything. Including also the ones that have simply been transformed from amcfast / apfelcomb.

[felixhekhorn]

Concerning DIS grids:

• can you compare the full metadata between the two? (including theory and obs card) they should match to the last digit (or else we have an indication on where to start).
• as a guess in the wild: this maybe sounds like a "theory card issue" - i.e. not the correct theory card has been used in one of the two cases?
• moreover a 20% diff is clearly not sufficient for a match
• also concerning @andreab1997 conjecture: this seems to affect only bins which lay in a 4FNS, correct? (because e.g. HERACC is not present and I believe this is all large Q2)

as for single top:

• these are converted grids, right? so there can not be a problem with the grids, correct? (the correct conversion might be an issue and to solve this we need a unification via a conversion backend https://github.com/NNPDF/pinefarm/issues/3 ) (the conversion argument also goes for the ttbar stuff, I guess)
• which leaves eko which wasn't an issue so far and is working for everybody else and
• pineko

[scarlehoff]

can you compare the full metadata between the two?

What do you mean? For theory 400 I have the fktables for Theory 424 I have the grids. What do you want me to compare exactly?

this maybe sounds like a "theory card issue" - i.e. not the correct theory card has been used in one of the two cases?

Yes. Which is also @andreab1997 conjecture.

as for single top:

The issue with singletop is mainly that the grids are the wrong shape. So yes, this is a conversion issue most likely.

[felixhekhorn]

What do you mean? For theory 400 I have the fktables for Theory 424 I have the grids. What do you want me to compare exactly?

pineappl info --show [grid or fktable] or, since actually to be fair the FK table has more metadata, pineappl info --get theory [grid or fktable] and pineappl info --get runcard [grid or fktable] as a start and then maybe some more fields ...

[scarlehoff]

For Theory 400:

{"CKM": "0.97428 0.22530 0.003470 0.22520 0.97345 0.041000 0.00862 0.04030 0.999152", "Comments": "NNPDF4.0 NNLO alphas=0.118", "DAMP": 0, "EScaleVar": 1, "FNS": "FONLL-C", "FactScaleVar": true, "GF": 1.1663787e-05, "HQ": "POLE", "IB": 0, "IC": 1, "ID": 200, "MP": 0.938, "MW": 80.398, "MZ": 91.1876, "MaxNfAs": 5, "MaxNfPdf": 5, "ModEv": "TRN", "NfFF": 4, "PTO": 2, "Q0": 1.65, "QED": 0, "Qedref": 1.777, "Qmb": 4.92, "Qmc": 1.51, "Qmt": 172.5, "Qref": 91.2, "RenScaleVar": true, "SIN2TW": 0.23126, "SxOrd": "LL", "SxRes": 0, "TMC": 1, "XIF": 1.0, "XIR": 1.0, "alphaqed": 0.007496252, "alphas": 0.118, "fact_to_ren_scale_ratio": 1.0, "global_nx": 0, "kDISbThr": 1.0, "kDIScThr": 1.0, "kDIStThr": 1.0, "kbThr": 1.0, "kcThr": 1.0, "ktThr": 1.0, "mb": 4.92, "mc": 1.51, "mt": 172.5, "nf0": null, "nfref": null}

{"PolarizationDIS": 0.0, "ProjectileDIS": "electron", "PropagatorCorrection": 0.0, "TargetDIS": "proton", "interpolation_is_log": true, "interpolation_polynomial_degree": 4, "interpolation_xgrid": [1.9999999999999954e-07, 3.034304765867952e-07, 4.6035014748963906e-07, 6.984208530700364e-07, 1.0596094959101024e-06, 1.607585498470808e-06, 2.438943292891682e-06, 3.7002272069854957e-06, 5.613757716930151e-06, 8.516806677573355e-06, 1.292101569074731e-05, 1.9602505002391748e-05, 2.97384953722449e-05, 4.511438394964044e-05, 6.843744918967896e-05, 0.00010381172986576898, 0.00015745605600841445, 0.00023878782918561914, 0.00036205449638139736, 0.0005487795323670796, 0.0008314068836488144, 0.0012586797144272762, 0.0019034634022867384, 0.0028738675812817515, 0.004328500638820811, 0.006496206194633799, 0.009699159574043398, 0.014375068581090129, 0.02108918668378717, 0.030521584007828916, 0.04341491741702269, 0.060480028754447364, 0.08228122126204893, 0.10914375746330703, 0.14112080644440345, 0.17802566042569432, 0.2195041265003886, 0.2651137041582823, 0.31438740076927585, 0.3668753186482242, 0.4221667753589648, 0.4798989029610255, 0.5397572337880445, 0.601472197967335, 0.6648139482473823, 0.7295868442414312, 0.7956242522922756, 0.8627839323906108, 0.9309440808717544, 1], "observables": {"XSHERANCAVG_charm": [{"Q2": 2.5, "x": 3e-05, "y": 0.11423}, {"Q2": 2.5, "x": 7e-05, "y": 0.11054}, {"Q2": 2.5, "x": 0.00013, "y": 0.09111}, {"Q2": 2.5, "x": 0.00018, "y": 0.091698}, {"Q2": 2.5, "x": 0.00035, "y": 0.054368}, {"Q2": 5.0, "x": 7e-05, "y": 0.15321}, {"Q2": 5.0, "x": 0.00018, "y": 0.15385}, {"Q2": 5.0, "x": 0.00035, "y": 0.11642}, {"Q2": 5.0, "x": 0.001, "y": 0.07763}, {"Q2": 7.0, "x": 0.00013, "y": 0.22486}, {"Q2": 7.0, "x": 0.00018, "y": 0.20231}, {"Q2": 7.0, "x": 0.0003, "y": 0.17669}, {"Q2": 7.0, "x": 0.0005, "y": 0.16158}, {"Q2": 7.0, "x": 0.0008, "y": 0.11994}, {"Q2": 7.0, "x": 0.0016, "y": 0.090227}, {"Q2": 12.0, "x": 0.00022, "y": 0.31613}, {"Q2": 12.0, "x": 0.00032, "y": 0.29041}, {"Q2": 12.0, "x": 0.0005, "y": 0.24098}, {"Q2": 12.0, "x": 0.0008, "y": 0.18134}, {"Q2": 12.0, "x": 0.0015, "y": 0.14761}, {"Q2": 12.0, "x": 0.003, "y": 0.10103}, {"Q2": 18.0, "x": 0.00035, "y": 0.31977}, {"Q2": 18.0, "x": 0.0005, "y": 0.29049}, {"Q2": 18.0, "x": 0.0008, "y": 0.25539}, {"Q2": 18.0, "x": 0.00135, "y": 0.20163}, {"Q2": 18.0, "x": 0.0025, "y": 0.163}, {"Q2": 18.0, "x": 0.0045, "y": 0.11367}, {"Q2": 32.0, "x": 0.0006, "y": 0.38846}, {"Q2": 32.0, "x": 0.0008, "y": 0.37557}, {"Q2": 32.0, "x": 0.0014, "y": 0.2807}, {"Q2": 32.0, "x": 0.0024, "y": 0.21897}, {"Q2": 32.0, "x": 0.0032, "y": 0.20149}, {"Q2": 32.0, "x": 0.0055, "y": 0.15534}, {"Q2": 32.0, "x": 0.008, "y": 0.094026}, {"Q2": 60.0, "x": 0.0014, "y": 0.32542}, {"Q2": 60.0, "x": 0.002, "y": 0.32893}, {"Q2": 60.0, "x": 0.0032, "y": 0.25762}, {"Q2": 60.0, "x": 0.005, "y": 0.1925}, {"Q2": 60.0, "x": 0.008, "y": 0.1596}, {"Q2": 60.0, "x": 0.015, "y": 0.094583}, {"Q2": 120.0, "x": 0.002, "y": 0.37661}, {"Q2": 120.0, "x": 0.0032, "y": 0.22743}, {"Q2": 120.0, "x": 0.0055, "y": 0.21729}, {"Q2": 120.0, "x": 0.01, "y": 0.15186}, {"Q2": 120.0, "x": 0.025, "y": 0.070217}, {"Q2": 200.0, "x": 0.005, "y": 0.23889}, {"Q2": 200.0, "x": 0.013, "y": 0.17035}, {"Q2": 350.0, "x": 0.01, "y": 0.223}, {"Q2": 350.0, "x": 0.025, "y": 0.10646}, {"Q2": 650.0, "x": 0.013, "y": 0.2026}, {"Q2": 650.0, "x": 0.032, "y": 0.088456}, {"Q2": 2000.0, "x": 0.05, "y": 0.060261}]}, "prDIS": "NC"}

---

For theory 424

{"CKM": "0.97428 0.22530 0.003470 0.22520 0.97345 0.041000 0.00862 0.04030 0.999152", "Comments": "NNPDF4.0 NLO alphas=0.118", "DAMP": 0, "EScaleVar": 1, "FNS": "FONLL-B", "FactScaleVar": true, "GF": 1.1663787e-05, "HQ": "POLE", "IB": 0, "IC": 1, "ID": 208, "MP": 0.938, "MW": 80.398, "MZ": 91.1876, "MaxNfAs": 5, "MaxNfPdf": 5, "ModEv": "TRN", "NfFF": 4, "PTO": 1, "Q0": 1.65, "QED": 0, "Qedref": 1.777, "Qmb": 4.92, "Qmc": 1.51, "Qmt": 172.5, "Qref": 91.2, "RenScaleVar": true, "SIN2TW": 0.23126, "SxOrd": "LL", "SxRes": 0, "TMC": 1, "XIF": 1.0, "XIR": 1.0, "alphaqed": 0.007496252, "alphas": 0.118, "fact_to_ren_scale_ratio": 1.0, "global_nx": 0, "kDISbThr": 1.0, "kDIScThr": 1.0, "kDIStThr": 1.0, "kbThr": 1.0, "kcThr": 1.0, "ktThr": 1.0, "mb": 4.92, "mc": 1.51, "mt": 172.5, "nf0": null, "nfref": null}

{"PolarizationDIS": 0.0, "ProjectileDIS": "electron", "PropagatorCorrection": 0.0, "TargetDIS": "proton", "interpolation_is_log": true, "interpolation_polynomial_degree": 4, "interpolation_xgrid": [1.9999999999999954e-07, 3.034304765867952e-07, 4.6035014748963906e-07, 6.984208530700364e-07, 1.0596094959101024e-06, 1.607585498470808e-06, 2.438943292891682e-06, 3.7002272069854957e-06, 5.613757716930151e-06, 8.516806677573355e-06, 1.292101569074731e-05, 1.9602505002391748e-05, 2.97384953722449e-05, 4.511438394964044e-05, 6.843744918967896e-05, 0.00010381172986576898, 0.00015745605600841445, 0.00023878782918561914, 0.00036205449638139736, 0.0005487795323670796, 0.0008314068836488144, 0.0012586797144272762, 0.0019034634022867384, 0.0028738675812817515, 0.004328500638820811, 0.006496206194633799, 0.009699159574043398, 0.014375068581090129, 0.02108918668378717, 0.030521584007828916, 0.04341491741702269, 0.060480028754447364, 0.08228122126204893, 0.10914375746330703, 0.14112080644440345, 0.17802566042569432, 0.2195041265003886, 0.2651137041582823, 0.31438740076927585, 0.3668753186482242, 0.4221667753589648, 0.4798989029610255, 0.5397572337880445, 0.601472197967335, 0.6648139482473823, 0.7295868442414312, 0.7956242522922756, 0.8627839323906108, 0.9309440808717544, 1], "observables": {"XSHERANCAVG_charm": [{"Q2": 2.5, "x": 3e-05, "y": 0.11423}, {"Q2": 2.5, "x": 7e-05, "y": 0.11054}, {"Q2": 2.5, "x": 0.00013, "y": 0.09111}, {"Q2": 2.5, "x": 0.00018, "y": 0.091698}, {"Q2": 2.5, "x": 0.00035, "y": 0.054368}, {"Q2": 5.0, "x": 7e-05, "y": 0.15321}, {"Q2": 5.0, "x": 0.00018, "y": 0.15385}, {"Q2": 5.0, "x": 0.00035, "y": 0.11642}, {"Q2": 5.0, "x": 0.001, "y": 0.07763}, {"Q2": 7.0, "x": 0.00013, "y": 0.22486}, {"Q2": 7.0, "x": 0.00018, "y": 0.20231}, {"Q2": 7.0, "x": 0.0003, "y": 0.17669}, {"Q2": 7.0, "x": 0.0005, "y": 0.16158}, {"Q2": 7.0, "x": 0.0008, "y": 0.11994}, {"Q2": 7.0, "x": 0.0016, "y": 0.090227}, {"Q2": 12.0, "x": 0.00022, "y": 0.31613}, {"Q2": 12.0, "x": 0.00032, "y": 0.29041}, {"Q2": 12.0, "x": 0.0005, "y": 0.24098}, {"Q2": 12.0, "x": 0.0008, "y": 0.18134}, {"Q2": 12.0, "x": 0.0015, "y": 0.14761}, {"Q2": 12.0, "x": 0.003, "y": 0.10103}, {"Q2": 18.0, "x": 0.00035, "y": 0.31977}, {"Q2": 18.0, "x": 0.0005, "y": 0.29049}, {"Q2": 18.0, "x": 0.0008, "y": 0.25539}, {"Q2": 18.0, "x": 0.00135, "y": 0.20163}, {"Q2": 18.0, "x": 0.0025, "y": 0.163}, {"Q2": 18.0, "x": 0.0045, "y": 0.11367}, {"Q2": 32.0, "x": 0.0006, "y": 0.38846}, {"Q2": 32.0, "x": 0.0008, "y": 0.37557}, {"Q2": 32.0, "x": 0.0014, "y": 0.2807}, {"Q2": 32.0, "x": 0.0024, "y": 0.21897}, {"Q2": 32.0, "x": 0.0032, "y": 0.20149}, {"Q2": 32.0, "x": 0.0055, "y": 0.15534}, {"Q2": 32.0, "x": 0.008, "y": 0.094026}, {"Q2": 60.0, "x": 0.0014, "y": 0.32542}, {"Q2": 60.0, "x": 0.002, "y": 0.32893}, {"Q2": 60.0, "x": 0.0032, "y": 0.25762}, {"Q2": 60.0, "x": 0.005, "y": 0.1925}, {"Q2": 60.0, "x": 0.008, "y": 0.1596}, {"Q2": 60.0, "x": 0.015, "y": 0.094583}, {"Q2": 120.0, "x": 0.002, "y": 0.37661}, {"Q2": 120.0, "x": 0.0032, "y": 0.22743}, {"Q2": 120.0, "x": 0.0055, "y": 0.21729}, {"Q2": 120.0, "x": 0.01, "y": 0.15186}, {"Q2": 120.0, "x": 0.025, "y": 0.070217}, {"Q2": 200.0, "x": 0.005, "y": 0.23889}, {"Q2": 200.0, "x": 0.013, "y": 0.17035}, {"Q2": 350.0, "x": 0.01, "y": 0.223}, {"Q2": 350.0, "x": 0.025, "y": 0.10646}, {"Q2": 650.0, "x": 0.013, "y": 0.2026}, {"Q2": 650.0, "x": 0.032, "y": 0.088456}, {"Q2": 2000.0, "x": 0.05, "y": 0.060261}]}, "prDIS": "NC"}

[andreab1997]

• also concerning @andreab1997 conjecture: this seems to affect only bins which lay in a 4FNS, correct? (because e.g. HERACC is not present and I believe this is all large Q2)

Actually the only points affected are the ones below the bottom threshold

[scarlehoff]

@andreab1997 looking at the runcard, it says PTO == 1, so it says that is is a NLO grid... but the grid is NNLO if I print the orders

pineappl obl HERA_NC_318GEV_EAVG_SIGMARED_CHARM.pineappl.lz4 --orders
o       order
-+-----------------
0 O()
1 O(as^1)
2 O(as^1 lf^1)
3 O(as^2)
4 O(as^2 lr^1)
5 O(as^2 lf^1)
6 O(as^2 lr^1 lf^1)
7 O(as^2 lf^2)

[felixhekhorn]

For Theory 400: For theory 424

huh? now I'm even more confused ... you're comparing an NNLO theory against an NLO one - they should never match! not even for some bins ... you're not comparing apples with apples and as said NLO is not a subset of NNLO

[felixhekhorn]

@andreab1997 looking at the runcard, it says PTO == 1, so this is a NLO grid... but the grid is NNLO if I print the orders

pineappl obl HERA_NC_318GEV_EAVG_SIGMARED_CHARM.pineappl.lz4 --orders
o       order
-+-----------------
0 O()
1 O(as^1)
2 O(as^1 lf^1)
3 O(as^2)
4 O(as^2 lr^1)
5 O(as^2 lf^1)
6 O(as^2 lr^1 lf^1)
7 O(as^2 lf^2)

actually no :upside_down_face: it is FONLL-B which fills some as^2 stuff but is considered NLO

[scarlehoff]

I'm saying 424 because that's the folder I was doing the calculation from... but they are the theory 200 grids... In theory.

In dom they are in /media/FK/test_pineline/data/grids/200/

[felixhekhorn]

I'm saying 424 because that's the folder I was doing the calculation from... but they are the theory 200 grids... In theory.

In dom they are in /media/FK/test_pineline/data/grids/200/

that might well be for the hadronic ones - for which NNLO does contain the NLO ones, but not the DIS ones

[scarlehoff]

Theory 200 is supposed to be NNLO also for DIS.

[felixhekhorn]

Theory 200 is supposed to be NNLO also for DIS.

it is! but 424 seems to be a NLO theory and so your comparing a NNLO theory (400) against a NLO theory (424) - @andreab1997 do you agree?

in pineko we're using PTO and not the orders of the grid (since indeed they are not reliable and we're doing a trick for FONLL-B (maybe I should not mention that detail :upside_down_face: ))

[scarlehoff]

Forget I said 424. I'm getting the grids from the Theory 200 folder.

So I'm guessing the theory 200 folder actually contains NLO grids which happen to be equal to the NNLO ones above the bottom mass https://github.com/NNPDF/runcards/issues/165#issue-1491712648 ?

[andreab1997]

@andreab1997 looking at the runcard, it says PTO == 1, so it says that is is a NLO grid... but the grid is NNLO if I print the orders

pineappl obl HERA_NC_318GEV_EAVG_SIGMARED_CHARM.pineappl.lz4 --orders
o       order
-+-----------------
0 O()
1 O(as^1)
2 O(as^1 lf^1)
3 O(as^2)
4 O(as^2 lr^1)
5 O(as^2 lf^1)
6 O(as^2 lr^1 lf^1)
7 O(as^2 lf^2)

Nono wait I believe you took the wrong theory card. The 424 theory card has for sure pto = 2

[andreab1997]

Theory 200 is supposed to be NNLO also for DIS.

it is! but 424 seems to be a NLO theory and so your comparing a NNLO theory (400) against a NLO theory (424) - @andreab1997 do you agree?

in pineko we're using PTO and not the orders of the grid (since indeed they are not reliable and we're doing a trick for FONLL-B (maybe I should not mention that detail 🙃 ))

Nono theory 424 is NNLO, what @scarlehoff is not the theory card of theory 424. I don't know what it is xd

[felixhekhorn]

Forget I said 424. I'm getting the grids from the Theory 200 folder.

fine - then wherever you got the two things before they can not match since they where computed with different settings ...

So I'm guessing the theory 200 folder actually contains NLO grids which happen to be equal to the NNLO ones above the bottom mass #165 (comment) ?

this I don't know, but NNLO corrections are active in all kinematic regions (most notably at small-x of course), so, as said, it is unlikely that this is the case

actually: @andreab1997 /media/FK/test_pineline/data/grids/200/ this might be the tutorial playground and it might be completely unphyiscal - so I'd be carefull there

now I'm convinced we're having a not-physics-issue, i.e. we just not using consistent files or settings

[scarlehoff]

I'm not taking any theory cards. I'm taking the grids in the Theory 424 folder which are a link from Theory 200.

pineappl info --get theory pineline_computation/data/grids/424/HERA_NC_318GEV_EAVG_SIGMARED_CHARM.pineappl.lz4 

{"CKM": "0.97428 0.22530 0.003470 0.22520 0.97345 0.041000 0.00862 0.04030 0.999152", "Comments": "NNPDF4.0 NLO alphas=0.118", "DAMP": 0, "EScaleVar": 1, "FNS": "FONLL-B", "FactScaleVar": true, "GF": 1.1663787e-05, "HQ": "POLE", "IB": 0, "IC": 1, "ID": 208, "MP": 0.938, "MW": 80.398, "MZ": 91.1876, "MaxNfAs": 5, "MaxNfPdf": 5, "ModEv": "TRN", "NfFF": 4, "PTO": 1, "Q0": 1.65, "QED": 0, "Qedref": 1.777, "Qmb": 4.92, "Qmc": 1.51, "Qmt": 172.5, "Qref": 91.2, "RenScaleVar": true, "SIN2TW": 0.23126, "SxOrd": "LL", "SxRes": 0, "TMC": 1, "XIF": 1.0, "XIR": 1.0, "alphaqed": 0.007496252, "alphas": 0.118, "fact_to_ren_scale_ratio": 1.0, "global_nx": 0, "kDISbThr": 1.0, "kDIScThr": 1.0, "kDIStThr": 1.0, "kbThr": 1.0, "kcThr": 1.0, "ktThr": 1.0, "mb": 4.92, "mc": 1.51, "mt": 172.5, "nf0": null, "nfref": null}

this I don't know, but NNLO corrections are active in all kinematic regions (most notably at small-x of course), so, as said, it is unlikely that this is the case

This is the weird thing .

[scarlehoff]

Ok. I will use the pinefarm to recompute all the DIS grids since I no longer trust theory 200, 424 or 400.

(Right now if I compute the grid from above with pinefarm I get a third result which is identical to theory 424/200 beyond the bottom mass, identical to theory 400 below and different from both for Q2=2.5)

[felixhekhorn]

Right now if I compute the grid from above with pinefarm I get a third result which is identical to theory 424/200 beyond the bottom mass, identical to theory 400 below and different from both for Q2=2.5

that is good! if you have a grid at Q2=2.5 involved (as I believe you do) then this is not reliable (and can be safely ignored), because there is no LHAPDF support in this region (it is below the fitting scale)

[andreab1997]

actually: @andreab1997 /media/FK/test_pineline/data/grids/200/ this might be the tutorial playground and it might be completely unphyiscal - so I'd be carefull there

now I'm convinced we're having a not-physics-issue, i.e. we just not using consistent files or settings

But in fact the place in which I compute stuff is not that. Is /media/FK/test_pineline/pineline_computation/. I don't believe we have produced unphysical grids just for the tutorial.

[scarlehoff]

that is good!

It's not good. It's schizophrenic :__

Anyway, I'm re-computing everything. I'm hoping that the theory 200 runcard in pinefarm is correct https://github.com/NNPDF/runcards/blob/master/theories/theory_200_1.yaml otherwise I will sit by the corner and start crying like only a grown man can cry.

[scarlehoff]

But in fact the place in which I compute stuff is not that. Is /media/FK/test_pineline/pineline_computation/. I don't believe we have produced unphysical grids just for the tutorial.

I'm afraid your grids are pointing to that folder

lrwxrwxrwx 1 andreab1997 nnpdf 86 Nov 17 10:59 pineline_computation/data/grids/424/HERA_NC_318GEV_EAVG_SIGMARED_CHARM.pineappl.lz4 -> /media/FK/test_pineline/data/grids/200/HERA_NC_318GEV_EAVG_SIGMARED_CHARM.pineappl.lz4

[andreab1997]

Ok. I will use the pinefarm to recompute all the DIS grids since I no longer trust theory 200, 424 or 400.

(Right now if I compute the grid from above with pinefarm I get a third result which is identical to theory 424/200 beyond the bottom mass, identical to theory 400 below and different from both for Q2=2.5)

Actually this is what I was doing before. What we found is that the Fk table of theory 400 can be reproduced (at least for HERACOMB_SIGMARED_C) using the grid with nf=4 but then using the eko produced with the theory card 424.yaml (which instead has nf=5). So it just means that theory 400 was constructed in an inconsinstent way and, as @felixhekhorn was saying` the problem was that we used two different theory cards. But we are able to reproduce it @scarlehoff

[andreab1997]

But in fact the place in which I compute stuff is not that. Is /media/FK/test_pineline/pineline_computation/. I don't believe we have produced unphysical grids just for the tutorial.

I'm afraid your grids are pointing to that folder

lrwxrwxrwx 1 andreab1997 nnpdf 86 Nov 17 10:59 pineline_computation/data/grids/424/HERA_NC_318GEV_EAVG_SIGMARED_CHARM.pineappl.lz4 -> /media/FK/test_pineline/data/grids/200/HERA_NC_318GEV_EAVG_SIGMARED_CHARM.pineappl.lz4

Yes but while the theory cards can be wrong, the grids are always the same: we have not produced unphysical grids just for the tutorial. We used the real ones. And from that folder I am only taking the grids.

[felixhekhorn]

using the grid with nf=4

in the yadism implementation of FONLL there is no concept of nf (and yadism is indeed ignoring it) so as long as you use FONLL you can put whatever value in that field (for @scarlehoff mental health sake I pray we really did that :upside_down_face: ) - that is not to say, that there still might be issues but these might as well be related to pineko

[andreab1997]

using the grid with nf=4

in the yadism implementation of FONLL there is no concept of nf (and yadism is indeed ignoring it) so as long as you use FONLL you can put whatever value in that field (for @scarlehoff mental health sake I pray we really did that 🙃 ) - that is not to say, that there still might be issues but these might as well be related to pineko

Wait, what I am trying to say is that my grids and the ones produced by pinefarm have different lumis. My grids have till the bottom while the ones produced by pinefarm have till the charm. Moreover, we are able to reproduce theory 400 only using the ones that have lumis till the charm.

[andreab1997]

using the grid with nf=4

in the yadism implementation of FONLL there is no concept of nf (and yadism is indeed ignoring it) so as long as you use FONLL you can put whatever value in that field (for @scarlehoff mental health sake I pray we really did that 🙃 ) - that is not to say, that there still might be issues but these might as well be related to pineko

Wait, what I am trying to say is that my grids and the ones produced by pinefarm have different lumis. My grids have till the bottom while the ones produced by pinefarm have till the charm. Moreover, we are able to reproduce theory 400 only using the ones that have lumis till the charm.

Problem is that with pinefarm I am not able to reproduce the ones that have lumis till the bottom, even if they are the ones I am using for sv

[scarlehoff]

After regenerating all DIS grids with pinefarm I get % agreement with Theory 400. This is not perfect but it is better than what I was getting before. This is grid Vs fktable so the differences are not due to pineko. I would be ok with this. I can believe that six month of changes can introduce a % difference.

Now the question is, @andreab1997 @felixhekhorn ... are the grids I regenerated correct? If I look at the lumis they seem nf=5

$ pineappl obl HERA_NC_318GEV_EP_SIGMARED.pineappl.lz4 --lumis
l      entry
--+-------------
0  1 × (-5, -11)
1  1 × (-4, -11)
2  1 × (-3, -11)
3  1 × (-2, -11)
4  1 × (-1, -11)
5  1 × (21, -11)
6  1 × ( 1, -11)
7  1 × ( 2, -11)
8  1 × ( 3, -11)
9  1 × ( 4, -11)
10 1 × ( 5, -11)

Going back to the SINGLETOP. Where can I find the grids for those @felixhekhorn ? These are just converted from nnpdf grids, but I want to have them all in one place so I can upload them to the repository.

[andreab1997]

After regenerating all DIS grids with pinefarm I get % agreement with Theory 400. This is not perfect but it is better than what I was getting before. This is grid Vs fktable so the differences are not due to pineko. I would be ok with this. I can believe that six month of changes can introduce a % difference.

Now the question is, @andreab1997 @felixhekhorn ... are the grids I regenerated correct? If I look at the lumis they seem nf=5

$ pineappl obl HERA_NC_318GEV_EP_SIGMARED.pineappl.lz4 --lumis
l      entry
--+-------------
0  1 × (-5, -11)
1  1 × (-4, -11)
2  1 × (-3, -11)
3  1 × (-2, -11)
4  1 × (-1, -11)
5  1 × (21, -11)
6  1 × ( 1, -11)
7  1 × ( 2, -11)
8  1 × ( 3, -11)
9  1 × ( 4, -11)
10 1 × ( 5, -11)

Going back to the SINGLETOP. Where can I find the grids for those @felixhekhorn ? These are just converted from nnpdf grids, but I want to have them all in one place so I can upload them to the repository.

This seems ok but have you tried to do the same for HERACOMB_SIGMARED_C?

[scarlehoff]

Yes.

pineappl obl HERA_NC_318GEV_EAVG_SIGMARED_CHARM.pineappl.lz4 --lumis
l     entry
--+------------
0  1 × (-5, 11)
1  1 × (-4, 11)
2  1 × (-3, 11)
3  1 × (-2, 11)
4  1 × (-1, 11)
5  1 × (21, 11)
6  1 × ( 1, 11)
7  1 × ( 2, 11)
8  1 × ( 3, 11)
9  1 × ( 4, 11)
10 1 × ( 5, 11)

[felixhekhorn]

Going back to the SINGLETOP. Where can I find the grids for those @felixhekhorn ? These are just converted from nnpdf grids, but I want to have them all in one place so I can upload them to the repository.

• when I converted them back at the time I stored them at /media/FK/fktables/data/appl_subgrids
• they should be a translation of https://github.com/NNPDF/applgrids using https://github.com/NNPDF/fktables/blob/main/convert_applgrids.sh (note that this script by @cschwan is outdated by now, since appl2pine is now part of the PineAPPL library)
• Note that the directory may contain too much grids

Concerning the cutting of bins: keep in mind that we had a dedicated function to act upon the instructions by the database and (I think) we decided actively to not repeat a fixed normalization value, but rely on the casting (contrary to the old implementation)

[felixhekhorn]

Now the question is, @andreab1997 @felixhekhorn ... are the grids I regenerated correct? If I look at the lumis they seem nf=5

for a grid this is fine - indeed when yadism acts in flavor space it considers flavors up to MaxNfPDF

After regenerating all DIS grids with pinefarm I get % agreement with Theory 400. This is not perfect but it is better than what I was getting before. This is grid Vs fktable so the differences are not due to pineko. I would be ok with this. I can believe that six month of changes can introduce a % difference.

• Mmm since there was barely any development in yadism I wonder whether % is acceptable ... is there some sort of pattern (as before)?
• do the theory and observable runcards match? (accessible as before from the metadata) how do the yadism versions compare?
• what exactly are you comparing: 400fk against new grid? or new fk? do the new fk compare well with the new grids? (i.e. eko is fine and out of the game)

[scarlehoff]

I'm comparing grids to Theory 400 FkTables. There is no obvious structure. The runcards (that I checked) are the same.