Check Fixed target DIS

[giacomomagni]

This PR is to implement the fixed target dis datasets from hepdata.

Summary

Experiment	Reproducibility	Note	TODO
BCDMS	✅	Two new observables, which superseede the old ones.
CHORUS	✅	Kinematics and central data match. Last point added manually. Minor change in Q2
EMC	✅	Kinematics and central data match. Branching ratio of 0.82 wrt to Hepdata. Add syst of 15%
NMC	🟡	The ratio dataset matches (except for uncertainties). ⚠️ $\sigma$ normalisation not clear for the P dataset. Two new observables, which are alternative to the old ones.
NUTEV	🟡	Implementation matches with Mason theses. Br and acceptance corrections where provided directly by the author. Minor change in Q2	acc_err not used ? where is the normalization error coming from ?
SLAC	🔴	Numbers are consistent with Phys.Lett.B 282 (1992) 475, but the source is unavailable. See comment below.

Common questions

• ⚠️ Old implementation used the breakdown of systematics when available. HepData use the combined one.
  How do we want to proceed? NO, only in the legacy data.
• ⚠️ Do we want to store the information about the old systematics somewhere? NO, only in the legacy data
• ⚠️ Do we want to store the old rawdata for SLAC ? NO, only in the legacy data

New variants

Dataset	hepdata	arxiv	Action	Motivation	Uncertainties	Variants
BCDMS_NC_NOTFIXED_P	link	link	no action taken, replaced by new observables	The old implementation (data and kinematic) coincides with the R=0, not averaged on $\sqrt{s}$ values. However the averaged and not averaged tables do not have the same kinematic.	Breakdown of syst. not given in HepData
BCDMS_NC_NOTFIXED_D	link	link	no action taken, replaced by new observables	The old implementation (data and kinematic) coincides with the R=0, not averaged on $\sqrt{s}$ values. However the averaged and not averaged tables do not have the same kinematic.	Breakdown of syst. not given in HepData
CHORUS_CC_NOTFIXED_PB	link	link	new variant	Kinematics and central value match	Breakdown of syst. not given in HepData	hepdata
EMC_NC_250GEV_DW	link	link	new variant	Kinematics and central match value	Old implementation has a 0.15 % syst not given in HepData. + nuclear u.	rzero
NMC_NC_NOTFIXED	link	link	new variant	Kinematics and central value match	Breakdown of syst. not given in HepData	hepdata
NMC_NC_NOTFIXED_P	link	link	no action taken, see new observables	Kinematics match, $\sigma$ was computed in the old buildmaster from R and F2	Full breakdown of systematics included in legacy.
NUTEV_CC_NOTFIXED_PB		link	new variant	Kinematics and central value match	copied from old buildmaster	hepdata with the updated BR and no nuclear uncertanties
SLAC_NC_NOTFIXED_P			no action taken. metadata amended.
SLAC_NC_NOTFIXED_D			no action taken. metadata amended.

New HEPDATA Observables

Observable	hepdata	arxiv	Motivation	Variants
BCDMS_NC_NOTFIXED_P_EM-F2-HEPDATA	link	link	Averaged values on $\sqrt{s}$ has different kinematics.	rzero, rqcd
BCDMS_NC_NOTFIXED_D_EM-F2-HEPDATA	link	link	Averaged values on $\sqrt{s}$ has different kinematics.	rzero, rqcd
NMC_NC_NOTFIXED_P_EM-F2-HEPDATA	link	link	Old implementation was $\sigma$, not $F_2$, breakdown of syst. not given
NMC_NC_NOTFIXED_D_EM-F2-HEPDATA	link	link	Not available before (possible double counting), breakdown of syst. not given

[giacomomagni]

Okay here we have the first interesting case:
central data and and kinematics (even if I swap sqrts with y) do not match exactly.
What do we want to do? New dataset ? New observable ? @scarlehoff @enocera

Other question:
do we really want to store the hepdata tables or maybe automatically download them ?

[enocera]

My opinion is as follows:

• implement a new data set, consistent with the information on Hepdata, given that the legacy data set is there, we could study later the impact of the "new" dataset w.r.t. the old;
• we want to download and store the tables.

[scarlehoff]

Should we do a complete new dataset or just have a legacy variant?

(this is a new situation due to the kinematics change so a decision has to be made!! I'm personally partial to have either a variant or an observable even if the legacy didn't come from the same paper)

[giacomomagni]

So the legacy and the hepdata version should come from the same reference.
I'd be in favor of creating a variant (with different kinematic, unc and data) so this way it will be clear that it's completely exclusive wrt to the legacy one, and they would belong under the same observable and dataset as they should.

But I'm not sure if kinematic variants are easy to support (or already working)

[scarlehoff]

But I'm not sure if kinematic variants are easy to support (or already working)

It might be working and if not it's very easy to support so no problem

[giacomomagni]

It might be working and if not it's very easy to support so no problem

But are you sure? for instance it will requires also different FKtables, ecc...

[scarlehoff]

Theory and data/uncertainties are supported by variants (at the beginning it was only for uncertainties but then we realized data and theory might need to change for these legacy cases).

We decided to create a new dataset if e.g. the number of points changed.

[scarlehoff]

https://www.slac.stanford.edu/pubs/slacpubs/5250/slac-pub-5442.pdf
https://www.slac.stanford.edu/pubs/slacreports/reports11/slac-r-357.pdf

[giacomomagni]

As agrred with @enocera we have added some modifications to the EMC and CHORUS data w.r.t the plain hepdata implementation.

• For Chorus. We have decided to retain in the hepdata variant also CHORUSISOTARGCOR and CHORUSQEDRADCOR, computed exactly as in the legacy variant. The first uncertainty is due the isoscalar target correction interpreted as uncertainty, while the secondo one takes into account possible QED effects (QED radiation correction interpreted as uncertainty).

• For EMC. We added to the hepdata implementation an addtional 15% "general" systematical uncertainty, which was taken as from buidmaster. We also take into account a 1.2% error from the Branching ration.

This should conclude the revision of these datasets. So the PR is ready for rewiev.

[scarlehoff]

Thanks @giacomomagni should we merge #2192 here before doing any checks?

[giacomomagni]

Thanks @giacomomagni should we merge #2192 here before doing any checks?

sure, we can do it!

[scarlehoff]

A quick look seems ok (I have to check the failing fitbot, it seems vp crashed?) but I have two questions:

1. Why is there no default variant? I would say that we want to use the dataset without having to specify variant: xx. Just select the variant that you consider the "most correct" and promote it to be the default.

1.b Shouldn't we have a variant: dw for some of the nuclear uncertainties or in those cases we just need to use the legacy?

2. Wouldn't it make more sense (for organizational purposes) to have the _HEPDATA datasets as a separate obaervable within the same folder as the others? I guess this is somewhat irrelevant, but given that the others are just a port I think they can live safely in the same folder.

RE the other PR, actually it might be possible to merge it directly to master so on second thoughts better finish it and merge the other by itself since it is basically harmless.

[giacomomagni]

A quick look seems ok (I have to check the failing fitbot, it seems vp crashed?) but I have two questions:

1. Why is there no default variant? I would say that we want to use the dataset without having to specify variant: xx. Just select the variant that you consider the "most correct" and promote it to be the default.

1.b Shouldn't we have a variant: dw for some of the nuclear uncertainties or in those cases we just need to use the legacy?

which dataset are you referring exactly?

2. Wouldn't it make more sense (for organizational purposes) to have the _HEPDATA datasets as a separate obaervable within the same folder as the others? I guess this is somewhat irrelevant, but given that the others are just a port I think they can live safely in the same folder.

This is not poissbile because the kinematic is not the same, or the number of points was different, see the tables at the top of the PR...

RE the other PR, actually it might be possible to merge it directly to master so on second thoughts better finish it and merge the other by itself since it is basically harmless.

okay

[scarlehoff]

which dataset are you referring exactly?

All of them, you left dataset_uncertainties empty for all of them (including the ones that are purely new)

This is not poissbile because the kinematic is not the same, or the number of points was different, see the tables at the top of the PR...

You can have two different observables with different kinematics. I'm not talking about variants. So you have BCDMS_NC_NOTFIXED_D_EM-F2 and BCDMS_NC_NOTFIXED_D_EM-F2-HEPDATA, both under BCDMS_NC_NOTFIXED_D.

[giacomomagni]

which dataset are you referring exactly?

All of them, you left dataset_uncertainties empty for all of them (including the ones that are purely new)

Okay sorry, let me fix it and rerun the fitbot.

You can have two different observables with different kinematics. I'm not talking about variants. So you have BCDMS_NC_NOTFIXED_D_EM-F2 and BCDMS_NC_NOTFIXED_D_EM-F2-HEPDATA, both under BCDMS_NC_NOTFIXED_D.

But here we're going in loop:

#2170 (comment)
#2170 (comment)

[scarlehoff]

Okay sorry, let me fix it and rerun the fitbot.

I'm not sure what the problem of the fitbot is

But here we're going in loop:

No, we just meant different things by dataset*, I think you meant a "set of data" while I meant really a dataset name that you would put in a runcard.
What I'm suggesting is having two datasets: BCDMS_NC_NOTFIXED_D_EM-F2 and BCDMS_NC_NOTFIXED_D_EM-F2-HEPDATA both part of the set BCDMS_NC_NOTFIXED_D.

*I know, the terminology has become impossible, probably the per-folder organization was not a good idea but now it's too late.

[github-actions]

Greetings from your nice fit 🤖 !
I have good news for you, I just finished my tasks:

• Fit Name: NNBOT-5e49a24cb-2024-12-04
• Fit Report wrt master: https://vp.nnpdf.science/jh6DywkkTDiQYOM-80xVdQ==
• Fit Report wrt latest stable reference: https://vp.nnpdf.science/es-mfMAcTSybbiSqmfyfAg==
• Fit Data: https://data.nnpdf.science/fits/NNBOT-5e49a24cb-2024-12-04.tar.gz

Check the report carefully, and please buy me a ☕ , or better, a GPU 😉!