Number of Direct Hits

A loose cut is made on the variable (Ndird(muon fit) - Ndird(cascade fit))/ Nhits at 0.14 in 2002 and 2003 and at 0.18 in the variable Ndird(muon fit)/Nhits in 2001.  The agreement between data and background MC is not good enough to include this cut in the Support Vector Machine, but this cut is useful to reduce  the amount of  data being dealt with.  The agreement in some of the other variables between data and MC is also improved after this cut is taken.
Taking Ndird(muon fit) - Ndird(cascade fit) rather than just Ndird(muon fit),as was done in 2001, improves separation between signal and background by bringing some signal events which happened to wind up with relatively large values of Ndird(muon fit) back into the signal spike, since the number of direct hits is rarely significantly larger for the wrong fit hypothesis than the right one. 

________: Real Data
- - - - - - - : Background MC
________: Signal MC

2001:

ndirdnhits.arrow.he.gif

2002: (Ndird(muon fit) - Ndird(cascade fit))/ Nhits

varplots/ndird.2002.gif

2003: (Ndird(muon fit) - Ndird(cascade fit))/ Nhits

varplots/ndird.2003.gif

home