Some JLIP systematic studies

1
Some JLIP systematic studies
Daniel Bloch, B. Clément, D. Gelé, S. Greder,
I. Ripp-Baudot (IReS Strasbourg)

• factorization of light quark mistags
• mistag systematics
• SystemD cross-checks on MC
• b-tag systematics

2
Factorization of mistags

• use the same fixed shape in all 4 ? ranges
• only the normalization is free
• factorization is valid for JLIP

3
Mistag systematics

• mistag efficiency e(neg tag) x Fhf x Fll
• e(neg tag) syst compare jes and emqcd data
  ? 5 per Et bin
• Fhf syst
  vary the c and b content in
  QCD MC by 10
  ? Fhf varies by 6 for
  c, by 8 for b
• Fll syst replace QCD MC by Z ?uds or Wjets,
  but no significant differences
• Overall systematics ?(5²6²8²) 11

4
Cross-checks of SystemD

• Aim test SystemD on Monte Carlo only
• MC sample used
  light QCD 4080
  
  (516k jets use highest pt track to mimic muons)
  c
  Z?cc?µ ttbar?c?µ (57k jets)
  b Z?bb?µ
  ttbar?b?µ (146k jets)
• Use 3 taggers" muon ptrel ? 0.7 GeV/c
  
• JLIP prob ?
  0.007 on muon-jet
• JLIP prob ? 0.05
  on opposite" jet
• Check the correlations between taggers
• Solve the system
  and compare results with
  expectations

5
8 equations / 8 unknowns
1 muon tag, 2 JLIP tag, 3 opposite tag
e tag-efficiency, a/ß correlation factors

• n nb ncl
• n1 e1b nb e1cl ncl
• n2 e2b nb e2cl ncl
• n3 e3b nb e3cl ncl
• n12 a12 e1b e2b nb ß12 e1cl e2cl ncl
• n23 a23 e2b e3b nb ß23 e2cl e3cl ncl
• n31 a31 e3b e1b nb ß31 e3cl e1cl ncl
• n123 a12 a23 a31 e1b e2b e3b nb
  .
  ß12 ß23 ß31 e1cl e2cl e3cl ncl

. . . . . . .
6
correlations for b-jets
a12
a23
a31
7
results / expected vs jet Et
ncl/n e1cl (muon) e2cl (JLIP) e3cl
(oppo)
nb/n e1b (muon) e2b (JLIP) e3b (oppo)
nb/n e1b (muon) e2b (JLIP) e3b (oppo)
8
b-tag systematics

• Using SystemD
• vary correlation coefficients within their (MC)
  errors 2.9
• vary the muon ptrel cut (from ?0.4 to
  ?1.0 GeV/c) 0.7
• ? 3 but take rather a conservative (?) 5
• Use this estimate and show it on many plots
  Et, ?, jet
  multiplicity,run number

9
SystemD results looks stable,
but difference with ptrel
fits not well understood