| negative chi2 in PndKinVtxFitter ! [message #12751] |
Wed, 26 October 2011 10:21  |
Albrecht Gillitzer
Messages: 49
Registered: March 2004
|
continuous participant |
From: *ikp.kfa-juelich.de
|
|
Hi
In the simulation (rev 13401) of pbar p --> Ds+ Ds-; Ds+ --> phi pi+, Ds- --> phi e- nuebar I found in some cases negative(!) chi2 values in the phi-->K+K- vertex fit, using PndKinVtxFitter. The occurrence is on the level of ~1% with respect to the number of all K+K- combinations, or even more than 5% with respect to the number of analyzed events. Sometimes extremely small positive chi2 values << 1 (e.g. 10**-5 or 10**-5) are seen which is also strange.
It also occurs if I fit the Ds+ and Ds- vertices with three charged tracks.
I then checked the macro for the CT TDR benchmark channel Ds+ Ds- in macro/run/tdrct/psi3770: it also occurs there. (I am now running the eta_c macro in macro/run/tdrct/eta_c, but I have no doubt it will also occur there.)
Usually people plot chi2 distributions only for positive values, so that this problem may have escaped our attention.
My question(s): has anybody else taken notice of this problem so far? Is someone willing to look into the PndKinVtxFitter code to find out where things go wrong? How much confidence can we have in a vertex fitter doing such things?
Albrecht
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| Re: negative chi2 in PndKinVtxFitter ! [message #12780 is a reply to message #12779] |
Tue, 15 November 2011 00:05  |
Marius Mertens
Messages: 55
Registered: January 2009
|
continuous participant |
From: *ikp.kfa-juelich.de
|
|
Hi Ralf,
Well, there was some nonzero chance for a tutorial or similarly helpful stuff to have escaped my attention. But I do get your point regarding documentation. I must admit that I got pretty lost when I tried to figure out how to transfer the information from object to object which convinces the fitter that the TCandidate it sees is a real (well, fake in this case) track which it can fit.
Thanks a lot!
Marius
|
|
|
|
GSI Forum
Search
Help
Members
Login
Home
