150 likes | 264 Views
Tau MET Corrections Update. Alfredo Gurrola (Texas A&M) JetMET Meeting: August 21, 2007. Tau MET Corrections (Update). Current Packages use corrected CaloJet information based on the jet response, CaloJet E T / GenJet E T (as a function of Jet E T and h) ,
E N D
Tau MET Corrections Update Alfredo Gurrola (Texas A&M) JetMET Meeting: August 21, 2007
Tau MET Corrections (Update) • Current Packages use corrected CaloJet information based on the jet response, CaloJet ET / GenJet ET (as a function of Jet ET and h) , to determine the corrections to Jets and MET • Basic goal is to use the information provided by Particle Flow (PF) to determine the proper way to make corrections to MET • PF t jet response & resolutions as a function of t ET and h • MET w/o and with t corrections • Determine the best way to implement the t MET corrections
Tau MET Corrections (Update) Tau MET Corrections (Update) • Let’s take a look at t PT resolutions! • CMSSW_1_6_0_pre4 • W tn sample of 90,000 events • |ht| < 2.5 for generator level taus • Compare PT resolutions for: 1. t jets reconstructed using IterativeCone5CaloJet algorithm 2. t jets reconstructed using IterativeCone5CaloJet algorithm with energy corrections based on jet response as a function of ET and h 3. t jets reconstructed using Particle Flow (PF) and tagged using RecoTauTag/PFConeIsolation Package which discriminates between t jets and QCD jets based on the narrowness of t jets NOTE: The MC t must be matched to BOTH the PF t and the InterativeCone5CaloJet
Tau MET Corrections (Update) • To make comparisons to MC t’s, a definition of a ‘match’ is required • Look at DR between PF t’s and MC t’s • Look at DR between Calorimeter Jets and MC t’s • Define a match between PF t’s and MC t’s : DR < 0.1 • Define a match between IterativeCone5 Calorimeter t jets and MC t’s : DR < 0.2 IterativeCone5 t Jets PF Jets tagged as t’s Jets that are NOT t’s
Tau MET Corrections (Update) • What is the t jet PT resolution for PF & how does it compare to the current jet resolutions (Calorimeter Jets & corrected CaloJets)? • Not only is the PT resolution better for PF, but the distribution is now peaked at 0! • Because of the longer tail in the PF distribution, RMS is NOT a good indicator of the PT resolution • Note: Distributions contain t’s of ANY visible energy & h (Statistics are dominated by low PT) IterativeCone5 t’s PF Corrected IterativeCone5 t’s
Tau MET Corrections (Update) • Since the statistics are dominated by low PT, the previous plots do not show the PT dependence. • Is there a PT dependence? IterativeCone5 t’s Corrected IterativeCone5 t’s PF
Tau MET Corrections (Update) TAIL!! • How do the resolutions compare at low PT? • Once again, we can see that PF has better resolution. • Tails of the PF distributions are caused by double counting in PF code! • Once these bugs are fixed, we should see less effect of the tails on the RMS IterativeCone5 t’s Corrected IterativeCone5 t’s PF
Tau MET Corrections (Update) • How do the resolutions compare at high PT? • At high PT, we really begin to see the huge differences in PF resolution • However, we still have the tails! IterativeCone5 t’s Corrected IterativeCone5 t’s PF TAIL!!
Tau MET Corrections (Update) • Let’s take a look at MET corrections for t’s using PF! • Event Selection • W tn sample • |ht| < 2.5 • Njets(QCD) = 0 with PT > 5 GeV • Correct Missing Energy if Monte Carlo t has PT > 15 GeV NOTE: The MC t must be matched to BOTH the PF t and the InterativeCone5CaloJet
Tau MET Corrections (Update) Before ME Corrections PF t ME Corrections Corrected CaloJets Missing Energy projected onto standard x axis and y axis • Missing Energy corrections for t’s using PF have better resolution • There is a HUGE difference in resolution between Missing Energy Corrections using PF t’s and Corrected IterativeCone5CaloJets • NOT much difference between Uncorrected ME and Corrected ME using PF. The reason is because we are dealing with low PTt’s (tails also affect resolution)
Tau MET Corrections (Update) TAIL!! • MET Resolution • Peak Position and resolution is better for MET corrections with PF t’s • The distribution for PF t MET corrections is narrower in the central parts of the distributions • However, the RMS is affected by the tails in the PF distributions (double counting) Before MET Corrections PF t MET Corrections MET Corrections with Corrected CaloJets
Tau MET Corrections (Update) • There is HUGE improvement between standard “jet based” MET corrections and PF t MET corrections • However, there is NOT much improvement between the distributions with PF t MET corrections and those without any MET corrections • The reason for this small (if any) improvement is that for W tn we are dealing with low PT taus • Let’s take a look at higher PT taus (larger MET) • H tn sample of 20,000 events • |ht| < 2.5 for generator level taus • Correct Missing Energy if Monte Carlo t has PT > 50 GeV
Tau MET Corrections (Update) IterativeCone5 t’s PT Resolution Corrected IterativeCone5 t’s PF
Tau MET Corrections (Update) • MET Resolution • Peak Position is better for MET corrections with PF t’s • Resolution for PF t MET corrections is worse than the resolution without t MET corrections, but better than the resolution for standard “jet based” MET corrections using IterativeCone5CaloJets Before MET Corrections MET Corrections with Corrected CaloJets PF t MET Corrections