Dear MG experts, As the title implies what I'd like to do is to recalculate the systematic uncertainties for scale and PDF variations after the event selection. So after running the embedded python tool for systematics, the original lhe file is enriched with a bunch of reweighted weights based the following indexing table as quoted in the enriched lhe file. MUR=0.5 MUF=0.5 MUR=0.5 MUF=0.5 dyn_scale_choice=sum pt MUR=0.5 MUF=0.5 dyn_scale_choice=HT MUR=0.5 MUF=0.5 dyn_scale_choice=HT/2 MUR=0.5 MUF=0.5 dyn_scale_choice=sqrts MUR=0.5 MUR=0.5 dyn_scale_choice=sum pt MUR=0.5 dyn_scale_choice=HT MUR=0.5 dyn_scale_choice=HT/2 MUR=0.5 dyn_scale_choice=sqrts MUR=0.5 MUF=2.0 MUR=0.5 MUF=2.0 dyn_scale_choice=sum pt MUR=0.5 MUF=2.0 dyn_scale_choice=HT MUR=0.5 MUF=2.0 dyn_scale_choice=HT/2 MUR=0.5 MUF=2.0 dyn_scale_choice=sqrts MUF=0.5 MUF=0.5 dyn_scale_choice=sum pt MUF=0.5 dyn_scale_choice=HT MUF=0.5 dyn_scale_choice=HT/2 MUF=0.5 dyn_scale_choice=sqrts dyn_scale_choice=sum pt dyn_scale_choice=HT dyn_scale_choice=HT/2 dyn_scale_choice=sqrts MUF=2.0 MUF=2.0 dyn_scale_choice=sum pt MUF=2.0 dyn_scale_choice=HT MUF=2.0 dyn_scale_choice=HT/2 MUF=2.0 dyn_scale_choice=sqrts MUR=2.0 MUF=0.5 MUR=2.0 MUF=0.5 dyn_scale_choice=sum pt MUR=2.0 MUF=0.5 dyn_scale_choice=HT MUR=2.0 MUF=0.5 dyn_scale_choice=HT/2 MUR=2.0 MUF=0.5 dyn_scale_choice=sqrts MUR=2.0 MUR=2.0 dyn_scale_choice=sum pt MUR=2.0 dyn_scale_choice=HT MUR=2.0 dyn_scale_choice=HT/2 MUR=2.0 dyn_scale_choice=sqrts MUR=2.0 MUF=2.0 MUR=2.0 MUF=2.0 dyn_scale_choice=sum pt MUR=2.0 MUF=2.0 dyn_scale_choice=HT MUR=2.0 MUF=2.0 dyn_scale_choice=HT/2 MUR=2.0 MUF=2.0 dyn_scale_choice=sqrts # scale # 262000: NNPDF3.0 LO global fit, alphas(MZ)=0.118. mem=0 ; average on replicas; mem=1-100 ; PDF replicas PDF=262000 MemberID=1 PDF=262000 MemberID=2 PDF=262000 MemberID=3 PDF=262000 MemberID=4 PDF=262000 MemberID=5 PDF=262000 MemberID=6 PDF=262000 MemberID=7 PDF=262000 MemberID=8 PDF=262000 MemberID=9 PDF=262000 MemberID=10 PDF=262000 MemberID=11 PDF=262000 MemberID=12 PDF=262000 MemberID=13 PDF=262000 MemberID=14 PDF=262000 MemberID=15 PDF=262000 MemberID=16 PDF=262000 MemberID=17 PDF=262000 MemberID=18 PDF=262000 MemberID=19 PDF=262000 MemberID=20 PDF=262000 MemberID=21 PDF=262000 MemberID=22 PDF=262000 MemberID=23 PDF=262000 MemberID=24 PDF=262000 MemberID=25 PDF=262000 MemberID=26 PDF=262000 MemberID=27 PDF=262000 MemberID=28 PDF=262000 MemberID=29 PDF=262000 MemberID=30 PDF=262000 MemberID=31 PDF=262000 MemberID=32 PDF=262000 MemberID=33 PDF=262000 MemberID=34 PDF=262000 MemberID=35 PDF=262000 MemberID=36 PDF=262000 MemberID=37 PDF=262000 MemberID=38 PDF=262000 MemberID=39 PDF=262000 MemberID=40 PDF=262000 MemberID=41 PDF=262000 MemberID=42 PDF=262000 MemberID=43 PDF=262000 MemberID=44 PDF=262000 MemberID=45 PDF=262000 MemberID=46 PDF=262000 MemberID=47 PDF=262000 MemberID=48 PDF=262000 MemberID=49 PDF=262000 MemberID=50 PDF=262000 MemberID=51 PDF=262000 MemberID=52 PDF=262000 MemberID=53 PDF=262000 MemberID=54 PDF=262000 MemberID=55 PDF=262000 MemberID=56 PDF=262000 MemberID=57 PDF=262000 MemberID=58 PDF=262000 MemberID=59 PDF=262000 MemberID=60 PDF=262000 MemberID=61 PDF=262000 MemberID=62 PDF=262000 MemberID=63 PDF=262000 MemberID=64 PDF=262000 MemberID=65 PDF=262000 MemberID=66 PDF=262000 MemberID=67 PDF=262000 MemberID=68 PDF=262000 MemberID=69 PDF=262000 MemberID=70 PDF=262000 MemberID=71 PDF=262000 MemberID=72 PDF=262000 MemberID=73 PDF=262000 MemberID=74 PDF=262000 MemberID=75 PDF=262000 MemberID=76 PDF=262000 MemberID=77 PDF=262000 MemberID=78 PDF=262000 MemberID=79 PDF=262000 MemberID=80 PDF=262000 MemberID=81 PDF=262000 MemberID=82 PDF=262000 MemberID=83 PDF=262000 MemberID=84 PDF=262000 MemberID=85 PDF=262000 MemberID=86 PDF=262000 MemberID=87 PDF=262000 MemberID=88 PDF=262000 MemberID=89 PDF=262000 MemberID=90 PDF=262000 MemberID=91 PDF=262000 MemberID=92 PDF=262000 MemberID=93 PDF=262000 MemberID=94 PDF=262000 MemberID=95 PDF=262000 MemberID=96 PDF=262000 MemberID=97 PDF=262000 MemberID=98 PDF=262000 MemberID=99 PDF=262000 MemberID=100 My first guess was that the systematics tool, based on these weights, calculates the systematic uncertainties as given on the terminal output that ends like this: INFO: #*************************************************************************** # # original cross-section: 3.32022e-07 # scale variation: +18.9% -15.1% # central scheme variation: +37% -5.41e-10% # PDF variation: +7.96% -7.96% # # dynamical scheme # 1 : 3.7871e-07 +20.1% -15.8% # \sum ET # dynamical scheme # 2 : 3.7871e-07 +20.1% -15.8% # \sum\sqrt{m^2+pt^2} # dynamical scheme # 3 : 4.54874e-07 +21.7% -16.7% # 0.5 \sum\sqrt{m^2+pt^2} # dynamical scheme # 4 : 3.32022e-07 +18.9% -15.1% # \sqrt{\hat s} First of all I'd like to ask if these variation percentages correspond to the standard error or to a confidence interval. If the second, at which CL? Secondly, the fact that the only seemingly symmetrical variation shown above is the PDF variation, drives me to think that, if not all of them, most of them are based directly on error propagation calculations (as derived from https://arxiv.org/abs/1110.4738) rather than the event (re-weighted) weights themselves. Is that the case? For the PDF variation, I tried calculating the standard error of the sum of re-weighted weights but the percentage, although close enough to the one printed out, is not identical. e.g. I got +-7.89%. Also, out of curiosity, I plotted quickly the sum of event weights (for 82500 events) for each PDF variation and I got the following histogram showing a distribution of the 101 PDF variations that declines a bit from a Gaussian. https://prnt.sc/p7m3v3 Can you confirm if that is expected? Finally, my main question is, how can I re-calculate these uncertainties based on the new weights after my event selection? I would like to avoid the hardcore way of removing the events from the lhe file and re-feeding it to the systematics tool. Thanks for bearing with me, Yannis

When	Who	Action	New State
2019-09-18	Olivier Mattelaer	Answer	Answered
2019-09-19	Yannis Maznas	Reopen	Open
2019-09-19	Olivier Mattelaer	Answer	Answered
2019-09-26	Yannis Maznas	Confirm	Solved

MadGraph5_aMC@NLO

History of question #684007