[Halld-physics] [Halld-offline] More Analysis Software Updates

[Matthew Shepherd]

Just to add a little to what Paul has written:

On Sep 26, 2014, at 6:25 PM, Paul Mattione <pmatt at jlab.org> wrote:

> With Matt's help, I've made some major improvements to the analysis software over the past few days (which aren't in the tagged release yet).  I've introduced "PreSelect" factories for DNeutralShower and DChargedTrack, which are responsible for removing "bad" tracks and showers before the analysis even starts: 

The idea here is to work towards standard quality + fiducial cuts that
all can use.  Eventually we need to be sure that the MC can
model the efficiency of these cuts using some study with 
actual data.  This will be a time-consuming validation process
and we don't want to have to repeat it or develop a different
technique for a wide variety of low-level track quality cuts.
It is nice to have a standard per-track systematic error.

> You can now optionally add this recommended cut to your plugin:
> 
> locReaction->Set_MaxExtraGoodTracks(4);
> 
> Which will cut events from your analysis right at the beginning that have too many extra "good" tracks.  "Good" is defined as "survives the "PreSelect" factory."  It's important to keep this value near 4 (for now) because there are still a fair number of ghost/junk tracks that masquerade as good tracks, and you don't want to throw away any signal events.  

I think we can show that this (and the corresponding 
statement for photons) is not as severe as GlueX 
folklore seems to imply.  The first piece of evidence is 
available in Ryan's talk at the software meeting:

https://halldweb1.jlab.org/talks/2014-3Q/SixFinalStates.pdf

See pages 12 and 13.  Here the blue histogram is the signal
and black points is all pythia.  Note that after "reasonable"
cuts on chi^2 very few events have more than 2 extra
tracks, even for 7-track topology 3(pi+pi-)p.  You can
see similar thing for photons, with states that have the
most charged tracks having the most "extra" photons
(presumably due to hadronic split-off showers).  One caveat
for Ryan's study is that a minimum 100 MeV cut was applied
on showers, which is probably a little more aggressive
than one wants in the BCAL.  

We're working to repeat this now for 30+ topologies
for the collaboration meeting.

I think it is important to remember that with GlueX you are
always trying to identify a particular reaction and almost
never studying inclusive reconstruction.  One should consider
this context when assessing the size of the background and
evaluate backgrounds only after some very basic
analysis cuts have been made to select a reaction.

Matt