Tag And Probe / BPAG Offline DQM


The purpose of this page is to document Muon HLT Offline DQM for tag and probe with muon resonances. The monitoring from Tag and Probe is supplemental to the baseline Muon HLT Offline DQM. The software was developed by the members of both the B PAG and the Muon HLT group. The resonances monitored are J/Psi, Upsilon, and Z.

What selection is applied to muons?

Probe track = tracker track
d0cut = 2.0
z0cut = 25.0  
RECO Muon Requirements: "pt > 0.5 && abs(eta) < 2.1 "
HLT Muon Requirements: "pt > 0.5 && abs(eta) < 2.1 "

Tag track = inner track
tagD0cut = 2.0
tagZ0cut = 50.0
tagRecoCuts = "pt > 1.0 && abs(eta) < 2.1 "
tagHltCuts  = "pt > 1.0 && abs(eta) < 2.1 "

Which triggers are monitored?

We monitor HLT_Mu3, HLT_Mu5 and HLT_Mu9 at each resonance.

How are the fits run?

  • We produce two invariant mass distribution for all tag and probe pairs in each bin of our kinematic variable of interest (pt, eta, phi, etc). The two distributions represent the numerator and the denominator of the efficiency measurement. The denominator is all RECO muon pairs, and the numerator is RECO muon pairs where the probe leg passes the trigger.
  • In each bin, we simultaneously fit the numerator and denominator invariant mass distributions. The fit scans through values of efficiency looking for a value of the efficiency that maximizes the likelihood. In pseudo code, this might look like:
         1. Start with an initial efficiency guess, e.g. 50%
         2. Fit Denominator and Numerator simultaneously with separate Guassian + Linear functions
            with the constraint the Numerator / Denominator = Efficiency
         3. Record the likelihood from the fit with that efficiency constraint.
         4. Try some other efficiency guesses near the original guess, and map how the fit likelihood 
             changes with efficiency
         5.  Use a gradient descent method to fit the value of efficiency that maximizes the likelihood
  • This choice of fit procedure has the benefit of a natural definition of error. The error on the efficiency is the 68% confidence band on the likelihood.
  • Example of a simultaneous fit of a numerator a denominator mass distribution in one bin of Pt:

Example Efficiency Plots

These efficiency plots were made by running on dimuon resonance samples.


J/PSI: 330_pre3 relval

Efficiency for the probe to pass HLT_Mu3



Upsilon: 330 relval

Efficiency for the probe to pass HLT_Mu5.



Z Mu Mu: 340_pre2 relval (partial sample)

Efficiency for the probe to pass HLT_Mu5


Where is the code?

-- JasonSlaunwhite - 2009-08-31

Topic attachments
I Attachment History Action Size Date WhoSorted ascending Comment
GIFgif tagProbeJPsi_HLT_Mu3.gif r1 manage 15.4 K 2009-10-23 - 17:00 JasonSlaunwhite  
GIFgif tagProbeUps_HLT_Mu5.gif r1 manage 15.3 K 2009-10-23 - 17:01 JasonSlaunwhite  
GIFgif tagProbeZ_HLT_Mu5.gif r1 manage 15.6 K 2009-10-23 - 17:03 JasonSlaunwhite  
GIFgif fitInPtBin5.gif r1 manage 23.4 K 2009-09-04 - 14:28 ZoltanGecse example fit

This topic: CMSPublic > CMSCommunicationsGroup > CMSCommunicationsProjects > WebHome > SWGuide > SWGuideOnSel > SWGuideTriggerStudiesHowTo > SWGuideMuonHLT > SWGuideMuonHLTOfflineDQM > SWGuideMuonHLTBPAGDQM
Topic revision: r4 - 2009-10-23 - JasonSlaunwhite
This site is powered by the TWiki collaboration platform Powered by PerlCopyright & 2008-2021 by the contributing authors. All material on this collaboration platform is the property of the contributing authors.
or Ideas, requests, problems regarding TWiki? use Discourse or Send feedback