[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: Fwd: Kfit with timing
Hall D PID Mail List:
Elke,
it seems there is a safety margin on the error understanding, since
doubling the errors still gives a reasonable BG suppression for p2K3pi
(BG goes from 10% to 30%), in MC with PYTHIA.
For very large (>100) suppression factors I would worry about flat backgrounds
and long tails.
BTW, my high energy experience was not outright positive with the error
handling. We tried to do our best with vertex separation in silicon u-strips
(resolution about 7 micron per hit). The residuals were never Gaussian
and the CL distributions were anything but flat. The role of multiple scattering
was modest (P>10 GeV), the errors were casued mainly by the pitch, pickup to the adjacent channels,
and, perhaps, delta-electrons. One may treat it as a superposition of several Gaussians,
with no way to separate them on the track to track basis. Perhaps, by now
someone has figured out how to treat such things consistently.
Eugene
------------------------------------------------------
Eugene Chudakov
http://www.jlab.org/~gen
phone (757) 269 6959 fax (757) 269 5703
Thomas Jefferson National Accelerator Facility
12000 Jefferson Ave, Newport News, VA 23606 USA
On Mon, 24 Mar 2008, Elke-Caroline Aschenauer wrote:
> On Mon, 24 Mar 2008, Matt wrote:
>
> Dear matt,
>
> what I'm saying is that at this stage of the experiment/tracking we have
> no idea how well we will be able to control our errors and systematics,
> such the suppression factors shown depend on this.
>
> I love MC but I'm sure we will have some effects we will have to include
> if we are running which we have no idea about now.
>
> So I just want to make sure we are careful wit hthe numbers we quote that
> is all.
>
> bye elke
>
>
>
>
> > Date: Mon, 24 Mar 2008 10:53:53 -0400
> > From: Matt <bellis@ernest.phys.cmu.edu>
> > To: Elke-Caroline Aschenauer <elke@jlab.org>
> > Cc: Alex Dzierba <dzierba@indiana.edu>, Eugene Chudakov <gen@jlab.org>,
> > halld-pid@jlab.org, cmeyer@ernest.phys.cmu.edu
> > Subject: Re: Fwd: Kfit with timing
> >
> > Hi Elke,
> >
> > The only thing which still worries me a bit is the assumptions on the
> > > error matrix for the tracking. I think we should show one or 2 example for
> > > the kinematic fits, but should clearly point out that this is still a
> > > point of concern.
> > >
> >
> >
> > I'm not sure I understand the point you keep trying to make when you say
> > "point of concern". You're basically saying if we don't understand our
> > errors, then the procedure doesn't work. I completely agree with you and
> > this was an issue with CLAS. A number of groups tried to build a consistent
> > kinematic fitter and failed because the error matrix returned from tracking
> > was wrong. It wasn't until Curtis and Mike Williams made a detailed,
> > systematic study of these errors that it started to work. In retrospect, the
> > study they did could've been done much earlier in CLAS' lifetime, but it
> > just wasn't given priority. One thing which I've emphasized from the first
> > talk I gave on this to the group last May, is that when we take data we will
> > need to have a concentrated effort to understand our error matrix as
> > returned by tracking.
> >
> > *Everything* we will do in GlueX will require a thorough understanding
> > of the systematics and errors of both the detector and simulation, and this
> > aspect of the analysis is no different. We shouldn't be trying to make these
> > first Monte Carlo studies with a broken error matrix, because we would not
> > expect to do analysis on real world data with a broken understanding of the
> > detector. It definitely warrants a bullet in any of these talks, not so much
> > to emphasize the problems, but to emphasize we know what we are doing and
> > that we understand what studies we will need to do in the early calibration
> > days of GlueX. The kinematic fit and it's usage should be considered no
> > different from any other aspect of our PID or reconstruction (pattern
> > finding, cluster algorithms) and we should expect as much effort to be put
> > into it's understanding as anything else in the experiment.
> >
> > The fact that the effective usage of the kinematic fit relies on our
> > understanding of the error matrix should be no more a "point of concern"
> > than any any other part of the reconstruction.
> >
> >
> > --
> > --
> > ----------------------------
> > Matt Bellis
> > Carnegie Mellon University
> > (office) 412-268-6949
> > (cell) 412-310-4586
> > ----------------------------
> >
>
> ( `,_' )+=-+=-+=-+=-+=-+=-+=-+=-+=-+=-+=-+=-+=-+=-+=-+=-+=
> ) `\ -
> / '. | +
> | `, Elke-Caroline Aschenauer =
> \,_ `-/ -
> ,&&&&&V Jefferson Lab +
> ,&&&&&&&&: HALL-D 12C / F381 121-A Atlantic Avenue =
> ,&&&&&&&&&&; Suite 8 Hampton, VA 23664 -
> | |&&&&&&&;\ 12000 Jefferson Ave +
> | | :_) _ Newport News, VA 23606 Tel.: 001-757-224-1216 =
> | | ;--' | Mail: elke@jlab.org Mobil: 001-757-256-5224 -
> '--' `-.--. | +
> \_ | |---' Tel.: 001-757-269-5352 =
> `-._\__/ Fax.: 001-757-269-6331 -
> +=-+=-+=-+=-+=-+=-+=-+=-+=-+=-+=-+=-+=-+=-+=-+=-+=-+=-+=-+=-+
>
>