# 2018-06-25 referees-meeting¶

Monday 25th June at 16:00 BST

Phone line:

UK Freephone: 08006948053

Caller PIN: 6783834683

https://www.intercallonline.com/listNumbersByCode.action?confCode=6783834683

## Discussion of Note¶

Discussion of the MICE Note MCSNote.pdf

## Notes¶

Present:

John Nugent

Chris Rogers

Paul Soler

Ken Long

John Cobb

Mariyan Bogomilov

- Done

- Moved now to Fig. 14; legend is still tiny

- Done
- Neither Moliere nor Carlisle-Cobb use radiation length; quote density as well.
- Please write "using the expression from the PDG"

- JHC: Why are there such fluctuations in the dx/dz MC plot? Is it a statistics issue? JN: There are 10x statistics in MC CR: there are three distributions on top of each other
- PS: Could consider the systematic discrepancy between MC and data in e.g. x'
- JHC: Define the selection before showing fig. 2

- Force latex to move the table to after Section 3.x
- Upstream track selection - duplication of 69.13 % looks like a bug; JN will check
- Diffuser cut text - define the radius
- Inconsistent r0 in text
- "45 degrees" -> explained in the text
- Move equations out of the table and into the text
- Note that the TOF cut for Table 3 has changed the statistics sample by factor 2

- Figure 3b should be a histogram
- Move to near S3.4 (and move fig. 13 forwards)
- Extrapolated distance is quite large; check number on envelope and add in MC
- Note that the statistics in 3c is rather small. Should enhance statistical size
- Add no hydride data

- Legend too small

KL: Plot chi2 MC vs data; consider adding chi2 selection

Fig. 5:- Doesn't show the selection cuts!; update so that it does actually show what the fiducial cut is
- Put scale in; e.g. relative to z = 0 m

- Add TOF12 resolution
- Show TOF12 momentum recon
- Why is 6d double peaked (before the correction)
- Fig. 6c has incorrect MAUS version stamp
- Fig. 6b points are at centre of each TOF bin;
- Add text to say what fig. 6b shows; put the formula in the text

- Fig. 8f why is there an event at x > 140 mm? JN: Cuts are made upstream, small probability of track scattering out of TKD
- Fig. 8d Why is there an event at r > 150 mm? JN: I don't know, but they get removed by fiducial; PS: tracker can make a track with r > 150 mm

- 12(a) Try moving by half a bin; is it a fluctuation or a feature?
- 11(a) A mild misalignment is noted; JN notes that the misalignment has been fixed.

- Use thicker markers; move before fig. 11 and 12

- Caption "generated" -> "calculated"; scattering distributions -> "theoretical scattering distributions"
- Should say MICE internal

- 15(a) what is going on at -40 mrad? Comparison with positive delta x of data and negative delta x of data
- Error bars may be underestimated; please review error bar calculation.
- Make asymmetry plots for "red curve" in Fig. 15

- Deconvolution is systematically not producing the correct answer
- JN: try using Gold's algorithm
- JN: used 10 iterations; not really a firm foundation for "10"
- KL: potential left-right asymmetry
- JHC: What about acceptance correction? JN: it makes the deconvolution worse so I took it out;
- JHC: Some events scatter in and then out; these will make the deconvolution imperfect; JN: use virtual planes on LiH to do the analysis (which does not have an acceptance associated); JHC virtual planes on LiH does not include effect of trackering scattering
- JN: Look at convergence of deconvolution;

JHC: Recommend abandoning deconvolution; publish on convolution only plus MC of data

PS: Wants to keep trying on deconvolution

- PS: would like a summary table of systematic errors

- JN: Disagreement probably arises due to deconvolution, not data

- JHC: Why are error bars so small? PS: the errors are statistically dominated; how does it look with statistical errors only?

- 3D distribution looks weird at low momentum
- Note asymmetry in 2D distributions

- JHC - should we really bother with cobb+carlisle? JN - well why not? we can discuss for the paper
- JN: if we abandon deconvolution don't we have to do forward convolution? JHC: no we just compare empty and full data with MC
- JN: try Gold's algorithm - couple of days work