[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Fwd: Kfit with timing



Hi Alex,
 
 
  My concern is, I'm sure shared by you,   that
we don't give the impression that we can compromise on
charged and neutral particle detection and three momentum
reconstruction and PID because we can employ kinematic fitting.
That will be another tool we need to identify signal over background
and have final four-vectors with which to do the analysis.

    I completely agree. I understand that there is concern about how best to present this material to show that we have considered these issues, but not oversell it as a universal panacea.

    The study is what the study is. David and I wanted to explore the idea of pi/K separation with kinematic fitting (and eventually TOF info) and we approached it with the simplest (but realistic) reactions we could use. The wrong way to approach this type of study with this software would have been to include a full pythia simulation and break events and add uncertainty to the error matrix, because then it would have been more difficult to disentangle all these effects from the actual performance of the fit results and attendant cuts/anticuts.   However, now that we've started here, all those other issues are fair game in the next round of studies.
 

On a somewhat related note -- if we have a 4pi p final state with
a cross section that is 10 x that for 2pi2K p (see Table 6) in the
overview document - how effectively does kinematic fitting remove
the latter reaction as a 10% background contribution for the former
without PID info except for TOF in the forward direction.  Another
way to put it is how well can we insure "pion purity".  Do you have
a feel for this?   You can imagine a resonant structure in the latter
reaction showing up as a 10% or so signal in the former as a
false peak after a PWA.


    This is discussed in the note. The analysis showed that under the conditions I imposed on the study, with a background 100x the signal, I could effectively remove that background entirely while only losing 25% of my signal events. And while some are surprised at that result, I would have been more surprised had I not got those numbers. This is 5 charged tracks, a 4C fit and 3-5 of those tracks have timing information associated with them. That is a huge number of constraints on the event when trying to make a mass determination.

    I didn't do the reverse in that particular study. That is, I didn't try to treat the  2pi2K as background to the 4pi events. But I would guess that I would see similar performance and cut it almost to 0.....under these conditions.

    I don't have a real feel of  what we'll see with a full background, realistic acceptance and other issues. But it points the directions for future studies as our simulation becomes more accurate and we get a better hold on our systematics. I think the idea of actually doing some limited strange channels in the first couple of years of running is not out of the question because of the great acceptance of the detector. But that is just my gut reaction and not based on any further studies...yet.

Matt B.



--
--
----------------------------
Matt Bellis
Carnegie Mellon University
(office) 412-268-6949
(cell) 412-310-4586
----------------------------