I'm thinking of implementing two kinds of thresholds for JrbCluster.
For each kind the user would be able to set separate values for had and em.
1. Ignore-threshold: If hit energy is below this threshold, disregard it
entirely.
2. Nucleus-threshold: If all hits in a cluster are below this threshold,
don't put it on the official cluster list. Instead, provide a method
which will return the list of such anemic clusters. A subsequent
analysis stage might decide to merge some of these clusters with
ones passing the cut.
While I'm at it, I'd like to take out the code that automatically
adds clusters to the event, and provide a method to do it explicitly
instead.
Advantages:
- flexibility in using refinement stages
- ability to run more than one clustering algorithm within
a job
Disadvantages:
- clumsier to use.
Comments?
Joanne
|