status of cc analysis
DESCRIPTION
Status of CC Analysis. J.Thomas for CC working group. Everything looks reasonable as function of pot and time. NUMI performance. Set of standard cuts agreed upon to select runs with ‘good’ beam. Set of loose cuts to get into MC/Data ball park. Data Preselection. Efficiency about 90% - PowerPoint PPT PresentationTRANSCRIPT
Status of CC Analysis
J.Thomas for
CC working group
NUMI performance
Everything looks reasonable as function of pot and time
Data Preselection
Set of standard cuts agreed upon to select runs with ‘good’ beam
Set of loose cuts to get into MC/Data ball park
Data PreselectionEfficiency about 90%
Purity about 80%
LE-10(170,185,200) pME pHE
5 special runs were taken to check out beam modelling
MC/Data disagreement changes with beam energy
Points to beam modelling, not detector modelling
PTxF
Different beams map out different areas of pT and xF.
Attempt to learn about secondary hadron production as well as beam settings from this informationBased on BMPT parametrization, try to fit it
SKZP Parametrization
SKZP Parametrization
First done for simple additional cuts on top of preselection
The results depend on the event selection.
The results shown are for the CC event selection described later
MC/Data Comparisons
Investigate the low level quantities: problem is always to make sure MC and Data are both apples before you compare themMake assumption that CC events better modelled/understood than NCEvent selection relies more heavily on muon modelling than shower modelling
SR fitter is used to track muons
SR Fitter Efficiency
Muon Tracking
Variables for CC selection
Variables for CC Selection
Agreement between MC and Data is not perfect at this stage
These variables are put into a PDF to select a high purity sample of CC events
PDF MC
Cut is imposed at –0.2 for ND and –0.1 for FD
MC/Data comparisons
Now that data and MC resemble each other, it is possible to compare low level quantities
Shower Ph/plane is NOT well modelled.
We have not seen this level of ‘dirty laundry’ from Super-K or K2K
EM modelling, reconstruction effects…..
Muon and Shower Energy
y and Total E Distribution
Intensity Studies
Special runs were carried out to study the effect of intensity on event selection
No non-linearities were observed
Very low intensity data proved this
Summary of data checks
The main worry is that the modelling agreement between Data and MC differs between ND and FD
There is no evidence that this is the case, even when MC and Data do not agree
Since Oxford, agreement between Data and MC has been significantly improved at low y by a reconstruction fixAn important cross check for all parameter measurement methods is to assume all ND data/MC difference is due to such problems
Oscillation Parameter MeasurementsThere are 4 methods currently being used to study our sensitivity
Fixing MC Generator parameters and systematics in the ND (DP)
F/N MC ratio to extrapolate ND data to FD predicted spectrum (TV)
GNUMI flux matrix to relate neutrinos of energy Ef in FD to neutrinos of energy En in near detector multiplies ND data to give predicted FD spectrum (NS)
Fitting of systematics, generator parameters in grid of E and y in ND (MI)
DP Fit
F/N extrapolation (TV)
This approach uses reconstructed energy dependent F/N to multiply contents of ND data energy bin to produce FD predicted spectrum
Oscillations are applied to Etrue in each bin of Ereco
MINUIT fitting allows exhaustive systematic studies to be carried out by means of 100s of ‘pseudo runs’ to estimate the combination of systematics at 1e20 statistics
F/N Pseudo Run : vanilla
F/N systematic checks
F/N Systematic Error
Matrix Method (NS)
This method uses the GNUMI flux files to generate a matrix which relates a neutrino of energy Ef in FD to one of energy En in ND
Ereco to Etrue in ND, cross section matrices cancel, efficiency of ND selection, matrix,FD efficiency,Etrue to Ereco via oscillations
Matrix Method Extrapolation
Matrix Method (NS)Systematic studies have been carried out on the sensitivity of this methodParticular systematics are applied to the FD or ND and FD spectra and then the fits carried out
Matrix Method (NS)
Data has 9% more events than MC in ND
m2 (eV2) sin2(2)
TRUE 0.002200 0.90
BEST FIT 0.002225 0.96
2006 MDC
A new MDC was carried out last week.
100 sub-runs of 1e20 were generated
100e20 data set was used to measure possible statistical biases in fit methods
All four methods were in agreement within errors
Second batch of MDC has been generated with combinations of systematics.
Some analysis have analised this too, but not all, not enough time
Systematics have been studied exhaustively in any case
F/N 2006 MDC
F/N 2006 MDC
Difference between pseudo runs and 100e20 is negligible in m2
Difference of 0.04 in sin22
Matrix 2006 MDCRight plots indicate the statistical fluctuations one might expect with 1e20 p.o.t
Left plot for 200e20
2006 MDC resultsDP method
m2 = 0.00225
sin22= 0.940
2=33.1/18
F/N
m2 = 0.0023+/-0.00006
sin22= 0.950 +/-0.027
2=64/57
Matrix
m2 = 0.00220+0.00005-0.00002
sin22= 0.96-0.025
2=43/12
Masaki
m2 = 0.00236
sin22= 0.933
2=33.1/18
Status of Cross Checks
F/N has been checked out by Niki
Matrix is being studied by Chris
David has been cross checked by Brian and Alycia
Masaki has not been cross checked
We will hear from Brian and Chris at this meeting
These cross checks will be completed before we go public
What now?
Exhaustive checks on data have been carried out10% statistical error suggests we are in a region of diminishing gainsSystematics are understood to be well below the statistical error
cc group proposal is to open the box tonightResults will be available tomorrow morningProgramme of cross-checks has to be put in place for the run up to the W&CProposal is for W&C to be on Thursday of Collaboration meeting
But first……
I’ve given a brief overview of where we are to put it all in context
Now we will listen in detail to the analyses
Questions should be addressed to the speaker as we go along
And finally, the FD spectrum
From ND data, predict the FD spectrum for no oscillation case
4 different methods give 4 different answers
ND selection the obvious culprit…