[Halld-physics] Massive Update to Kinematic Fitter & Analysis Library

Paul Mattione pmatt at jlab.org
Tue Apr 15 15:36:04 EDT 2014

I've just checked into the trunk a massive update to the kinematic fitter and analysis library.  Mike and I have tested these changes on a number of channels and things appear to be working, but I couldn't test all cases so please let me know if you have any problems.  The plan is to have plots from these studies ready to be shown at the next physics meeting.  

Full Documentation: https://halldweb1.jlab.org/wiki/index.php/GlueX_Analysis_Software


1) Improved convergence in vertex-p4 fits.  It turns out that this problem wasn't as large as I originally thought, so the improvements are very slight (a few %) except for a few extreme cases (e.g. the pi- from the cascade decay is missing).  

2) Added support for disabling specific mass constraints (set in DReactionStep).  This is useful for when you want to study/know the background underneath one of the mass peaks.  

3) Added support for inclusive kinematic fits.  If your DReaction contains a missing unknown particle or starts with a unknown decaying particle, you can now kinematically fit the reaction.  It will not apply a p4 constraint, but will still apply any other mass & vertex constraints.  

4) Tweaked interface for creating/setting constraints in manual kinematic fits.  See the wiki documentation for details. 

5) Confidence Level histogram now shows details about the fit: # of constraints, unknowns, and degrees of freedom, as well as exactly what constraints were applied (e.g. what particles were constrained to what vertex (vertex-constraining particles are blue)).  

6) Various other tweaks & bug fixes. 


1) New constructors have been added to DHistogramAction_MissingMass(Squared) and DCutAction_MissingMass(Squared) that allow you to histogram/cut on the missing mass (squared) off of a particle or a set of particles. 

2) Thrown/reconstructed particle matching now uses the tracking/shower covariance matrix for computing the match FOM.  This match FOM is also now stored along with the match itself.  It can be cut on when filling your ROOT TTree (set in DReaction) so that only "good" matches are marked in the ROOT TTree (although I recommend not cutting for now (default)).  Several histogram actions now include a (modifiable) cut (default +/- 5 sigma) were matching is used (e.g. determining PID in DHistogramAction_DetectedParticleKinematics()).  

3) A new action has been created that will allow to cut on whether or not the thrown track topology matches your DReaction: DCutAction_ThrownTopology.  In the constructor you can also specify whether or not you require an exact match, or that your DReaction be a subset of the thrown topology.  This is useful for determining whether an event has the specific particle decay chain you are looking for.  Note that you can also directly call the DAnalysisUtilities::Check_ThrownsMatchReaction() function instead. 

4) Various other tweaks & bug fixes. 

 - Paul

More information about the Halld-physics mailing list