Final revision of the 4.0 dataset

[scarlehoff]

As announced in the email, since we are close to the end, this is the assigment / organization of the datasets left.
This (more or less) matches what's in the wiki at the bottom of https://www.wiki.ed.ac.uk/display/nnpdfwiki/NNPDF4.1+Data+and+Theory+implementation+status

Dataset	PR	Implementation	Review
HERA data	#2175	@peterkrack	@t7phy
DYE866_Z0_800GEV_DW_RATIO_PDXSECRATIO	#2248	@peterkrack	@Radonirinaunimi @enocera
DYE866_Z0_800GEV_PXSEC	#2248	@peterkrack	@Radonirinaunimi @enocera
DYE605_Z0_38P8GEV_Z0_38P8GEV_DW_PXSEC	#2248	@peterkrack	@Radonirinaunimi @enocera
DYE906_Z0_120GEV_Z0_120GEV_DW_PDXSECRATIO	#2248	@peterkrack	@Radonirinaunimi @enocera
ATLAS_Z0_7TEV_LOMASS_M	#2171	@comane	@jacoterh
ATLAS_Z0_7TEV_46FB_CC-Y	#2237	@ecole41	@jacoterh
ATLAS_Z0_7TEV_46FB_CF-Y	#2237	@ecole41	@jacoterh
ATLAS_WPWM_7TEV_36PB_ETA	#2223	@ecole41	@jacoterh
ATLAS_Z0_7TEV_36PB_ETA		@ecole41	@jacoterh
ATLAS_Z0_8TEV_LOWMASS_M-Y	#2205	@comane	@t7phy
ATLAS_Z0_8TEV_HIMASS_M-Y	#2206	@comane	@t7phy
ATLAS_WPWM_13TEV_TOT	#2207	@comane	@achiefa
ATLAS_Z0_13TEV_TOT	#2207	@comane	@achiefa
CMS_Z0J_8TEV_PT-Y	#2241	@achiefa	@RoyStegeman
CMS_WCHARM_7TEV_WPWM-{RATIO, TOT}	#2245	@achiefa	@RoyStegeman
CMS_WCHARM_13TEV_WPWM-TOT-UNNORM	#2244	@achiefa	@RoyStegeman

I've tried to match the datasets among the people who volunteered/was in the wiki such that either the same person reviews the same dataset (e.g., ATLAS_Z0_7TEV_46FB_CC-Y and ATLAS_Z0_7TEV_46FB_CF-Y) or datasets that they already got experience with (e.g., @t7phy with HERA, @jacoterh and @achiefa DY, etc).

The FTDY might be a bit tricky in the same way as the FTDIS so I put two of you @Radonirinaunimi @enocera

The way I've been checking the datasets is by using the following runcard. I usually run it first with master to have a benchmark report, then with the branch (removing the variant legacy if needed). If the result of {@ total_chi2_data @} is exactly the same, that means that data and covmat are exactly the same as well. You can also add some other dataset from the same experiment to check that the correlation works.

check.yml

With more or less the following checklist:

• Is the chi2 the same between the old and new implementation?
• Do the theory-comparison plot look the same (labels and plot themselves) between master and the branch?
• Is the kinematics_override in the metadata.yaml set to identity?
• Do the variables for the dataset and their labels make sense?
• Does the xQ map look the same or, if it changed, does it look sensible?

fwiw, I've only looked superficially at the filter files. I've only looked more in detail when something in the list above seemed fishy.

meta:
  title: Data vs Th
  keywords: comparison
  author: me

pdfs: [NNPDF40_nnlo_as_01180]

pdf: NNPDF40_nnlo_as_01180

theoryid: 40_000_000

use_cuts: "internal"
marker_by: "dataset"

dataset_inputs:
    - {dataset: <name of the dataset>< , variant: <legacy when running with master>}

template_text: |

  Data-TH comparison
  ------------------
  {@ dataset_inputs plot_fancy @}

  Normalized
  ----------
  {@ dataset_inputs::pdfs plot_fancy(normalize_to=data)@}

  xq mapping
  ----------
  {@ plot_xq2 @}

  Chi Data
  --------
  {@ experiments_chi2_table @}
  {@ total_chi2_data @}


actions_:
  - report(main=true)

I think this is also a good opportunity for more people to gain some confidence with reviewing stuff so that it doesn't always fall on the same four people (not sure why that is, maybe fear of singing off on broken things... we do that all the time, it is not the end of the world!).

We almost finished! It has only taken 3 years!!111 #1436