Uncertainties of scatter variables account for hit efficiency

[GilesStrong]

Current state

PanelScatterBatch computes hit uncertainties as 1/(resolution*efficiency). These uncertainties are considered when fitting trajectories in AbsScatterBatch.get_muon_trajectory; relative weight of hit increases if resolution or efficiency increases, which should help with optimisation.

Potential problem

AbsScatterBatch._compute_out_var_unc computes the uncertainty on variables computed from the trajectories, but does so considering the uncertainty on the hits (which includes the efficiency). Is this a problem? Since efficiency will always be < 1 this leads to an artificial and unphysical increase in the uncertainties, but could be useful from an optimisation point-of-view...

Potential changes

I think that we should either:

1. Switch to computing hit "uncertainties" using only the hit resolution, and use this when computing the variable uncertainties. When performing track fitting, we can still include the efficiency in the weight.
2. Make it explicit that the "uncertainty" on scatter variables is not the actual uncertainty, but a combination of uncertainty and efficiency.

To avoid assumptions on downstream inference and other usage of the scattering variables, my preference would be for change 1.

[giamman]

Hello, I strongly prefer option 1. Whenever we intend to make efficiency part of the optimization, better to use it as an external term.

[vischia]

Hello, me too.

Cheers, Pietro

On Wed, 26 Oct 2022 at 11:42, Andrea Giammanco @.***> wrote:

Hello, I strongly prefer option 1. Whenever we intend to make efficiency part of the optimization, better to use it as an external term.

— Reply to this email directly, view it on GitHub https://github.com/GilesStrong/mode_muon_tomography/issues/141#issuecomment-1291772887, or unsubscribe https://github.com/notifications/unsubscribe-auth/ABAENKC5TRHORSI7ZCIXMMTWFD4JLANCNFSM6AAAAAAROXT35Y . You are receiving this because you are subscribed to this thread.Message ID: @.***>

-- Please note that, regardless of when I sent this email, I don't expect you to answer me outside of the regular working hours of your local time zone. Dr. Pietro Vischia FNRS Fellow http://cp3.irmp.ucl.ac.be/Members/pvischia mobile: 0034 666 986 616 (ES) office: +32 10 473205 mail: @. @. skype: rambling_hadron