I am going to include a slide on the kinem. fitting with the something
like:
- 3-C fit improves the track resolutions
- kin. fitting allows a considerable improvement in identification of
events with kaons over simple kin. constraints.
- the results depend on prominance of backrounds which are flat
or non-Gaussian in the residual space.
- in the ideal case a strong possible suppression can been obtained
- the work is in progress
I think that is mostly OK, but I don't know why you say 3-C fit. Curtis has a slide in his talk (and a section in his overview document) showing how much a 1-C fit (1 missing particle) improves the resolution for a invariant mass peak, using the standard kinematic fitter checked into the Hall-D repository. I'm not even sure why you would ever do a 3-C fit. I understand you have now written your own fitter, but the standard kinematic fit uses a 4-C fit for exclusive reactions. To be honest, if you say 3-C fit, it sounds like we don't know as much what we're doing or that we haven't put as much development into the fitter as we have.
You may send me a few statements (items) you would like to make.
Also, you may include a plot. Will you do it?
Yes, I'll put together a write up of what I think are the pertinent points and get it to you.
Yesterday, I tried the 3-C fit, using the event 3-momentum. It gave me some additional
BG reduction, a factor of 2 perhaps, with simple cuts and without
a signal suppression. I will look further into it.
If you are going to start using kinematic fitting, you should use the one checked into the Hall-D software repository. We have too much to do over the next year to start reinventing the wheel. If there are problems with my fitter they will be discovered quicker if more people use it, rather than everybody developing their own fitter. I understand that because we are under a time crunch that it may have been quicker for you to write your own fitter than learn the standard one, but I don't think it serves us well to have this parallel development.
BTW, what improvement in track momentum resolution did you obtain with the
3-momentum fit? For p2K3pi events I get on average 2.4%-->2.0% - not that much.
I assumed the diagonal covariance matrix for p,theta,phi.
The energy residual becomes very narrow for "true" events,
so adding the 4-th fit should not change the track resolution.
The improvement in momentum resolution is of course dependent on the kinematics, but it is probably on that order....maybe a little better. I think a better measure of the improvement is invariant mass peaks, as Curtis has shown. I too use a diagonal covariance matrix in p/theta/phi, though the fitting is done in x/y/z where the error matrix is not diagonal.
I'll get some slides to you and you can choose what you think is appropriate for the talk.
Matt