Can we be fooled at 1e20?

Jenny Thomas

UCL

PreviewExplanation of Quick Fit Method

Potential pitfallsStatistics

P.O.T

Calibration

FD MC/Data

Conclusion

Explanation of MethodBasic ‘know nothing’ approach

Use ND data in combination with MC to get predicted, unoscillated spectrumEncompass ND data/MC discrepancy as error bars in predicted spectrumAdd in beam uncertainty estimate for uncertainty in acceptanceUse simple 2 minimization with MINUIT to see effect on parameter measurement

Point is an independent cross-check of parameter measurement and useful tool for systematic studies

Explanation of MethodThis is an overestimate of errors if:

Due to detector modellingLight in scintillatorGEANT

Beam related (different from other)

This is reasonable ifND/FD different

SlicingCalibrationP.O.T

This could be an underestimate ifFD MC/Data totally different

Explanation of MethodAssumes all oscillated to

Etrue is oscillated and returned to Ereco

NC and CC events in sample but cuts applied to reduce NC

Pseudo runs study effect of statistics:Nearest integer of

#+ran(sqrt(#) + #+ran(sqrt(#))

For each bin in Ereco is calculated

Example of results

Potential Pitfalls at 1e20Using 100 pseudo-experiments

MINUIT calculates errors on parametersCan check these are reasonable

Potential Pitfalls at 1e20

0.0022eV2

0.005eV2

Protons on TargetErrors in #P.O.T can give false information: fitting within limits as an extra parameter solves this

Plots to right have a 10% P.O.T miscalibration

Calibration +5% Difference between ND and FD calibration can be studiedEffects worse at low m2

Again, extra parameter can alleviate this problem

Conclusion Effects of calibration and normalization should not hurt us at 1e20

FD Data/MC differences are the only comparison which isn’t taken into account in this method

At what value of m do we decide to raise the beam energy?