DUNE Fitter Validation Daniel Cherdack Colorado State University DUNE LBPWG Meeting Monday July 3 rd , 2017 1
Motivation ● DUNE now has access to several Fitters – GloBES/MGT – LOAF – VALOR – Cafana – Lfit ● Major benefit of many fitters is that they should validate each other ● Plan: ensure validation though a series of fitter tests that check each step in the fitting process ● Notes: – This is a proposal that will be refined; all specific are up for debate – So far plan is for FD only fits – Going forward all public plots must be from ‘approved’ fitters 2
Define the Baseline ● Compare all fitters to GloBES – Most publicly used plots came from GloBES – Already the unofficial method ● Setup/Inputs: – FD MC MMC7 with MVA cuts at 0.8 – 40 kt fiducial mass – 10 yrs @ 1.07 MW – Standard 1300 km baseline, and constant earth density used in CDR – Full CDR systematic treatment – Ability to propagate shape changing systematics 3
Spectra Matching ● Produce the 4 ( ν µ ,/ ν µ , ν e / ν e ) oscillated reconstructed energy spectra – Bins of 250 MeV from 0 to 10 GEV – Nufit v3 best fit osc param values (http://www.nu-fit.org/?q=node/139), and – δ CP = [- π /2, 0, π /2] ● Spectra do not have to be identical, but should agree with some reasonable (tbd) margin of error. ● Produce tables of bin contents by channel for easy comparison – ν e / ν e appearance – ν µ ,/ ν µ disappearance – ν τ ,/ ν τ appearance – Intrinsic ν e / ν e – NC 4
Systematic Error Propagation ● Choose a set of systematic uncertainty parameters, examples: – A normalization parameter like the NC norm – A flux throw from the flux covariance matrix – A GENIE reweight parameter like M ares – A GENIE FSI parameter – An energy scale parameter ● Produce the a set of spectra and tables for each parameter – Similar to those propsed on previous slide – Use only δ CP = 0 – Set parameter values to +1 σ (or ±1 σ ) 5
χ 2 Calculations ● Choose a few points in parameter space, examples: – The nominal MC – The GLoBES BF point – Every parameter at +1 σ ● Calculate the χ 2 ● Break down the contributions from – Each of the 4 spectra – The penalty term for each parameter ● Assume a CDR configuration – CDR Volume 2, page 3-44, top paragraph and table 3.9 – http://lbne2-docdb.fnal.gov/cgi-bin/RetrieveFile?docid=10688&filename =DUNE-CDR-physics-volume.pdf&version=10 6
χ 2 Minimization ● Fit for CPV sensitivity with true δ CP = [- π /2, 0 , π /2], and report: – Best fit parameter values – 1 σ uncertainty ranges for each fit parameter – Including nuisance parameters ● Fit a Mock Data samples – TBD, but something other than the Asimov data – Report the same results. ● Assume a CDR configuration – CDR Volume 2, page 3-44, top paragraph and table 3.9 – http://lbne2-docdb.fnal.gov/cgi-bin/RetrieveFile?docid=10688&fi lename=DUNE-CDR-physics-volume.pdf&version=10 7
CPV Sensitivity Plots ● Produce the standard set of CPV plots ● Assume a CDR configuration – CDR Volume 2, page 3-44, top paragraph and table 3.9 – http://lbne2-docdb.fnal.gov/cgi-bin/RetrieveFile?docid=10688 &filename=DUNE-CDR-physics-volume.pdf&version=10 8
Producing the GLoBES Baseline (per Elizabeth) ● Spectra matching: Ready ● Systematic error propagation: – Can do norm/CDR style systematics – Requires MGT for more complex systematics ● χ 2 Calculations: – No existing code to do this, but should not be hard to implement – Penalty terms for each parameters may be difficult (might be in MGT) ● χ 2 Minimization – Similar issues reporting nuisance parameter values and errors (might be in MGT) ● CPV Sensitivity: Ready 9
Recommend
More recommend
