100 likes | 282 Views
Systematic studies. Daniel Bloch (IReS Strasbourg). ptrel fits vs SystemD SystemD systematics mistag systematics: still to be studied. ptrel fits rely on ptrel templates. ptrel fits weakness.
E N D
Systematic studies Daniel Bloch (IReS Strasbourg) • ptrel fits vs SystemD • SystemD systematics • mistag systematics: still to be studied D. Bloch (IReS Strasbourg)
ptrel fits rely on ptrel templates D. Bloch (IReS Strasbourg)
ptrel fits weakness • The light template is estimated by taking any track in jet trigger data: it is a first approximation, but how to describe reconstruction pbs due to decay-in-flights, the pion/kaon weights, the change of direction between muon and primary pion/kaon ? • Which uncertainty on the b->mu, b->c->mu, c->mu MC templates ? • If the c/light ratio is fixed, what is its error and how does it depend on the muon or jet pt ? • Are the ptrel templates eta dependent ? • How to take into account the MC statistics in the b and c templates (it is not trivial at all) ? • Is there a correlation between the ptrel shapes and the lifetime tagging cut ? D. Bloch (IReS Strasbourg)
SystemD basics • Relies on 2 taggers (JLIP and muon ptrel) and 2 samples (muon-in-jet n and subsample p with opposite JLIP tagged jet) • Allows to solve 8 equations with 8 unknowns: D. Bloch (IReS Strasbourg)
SystemD assumes that: • the correlation between JLIP tag and ptrel cut is small: estimated on MC band clfactors • the c/light ratio is the same in the full sample and in the opposite tag subsample:factor, varied around 1on real data • the b-tag efficiency is the same in the full sample and in the opposite tag subsample:factor estimated from b MC • ptrel cut has the same efficiency for c and light:estimated by varying the ptrel cut on real data and that’s all ! D. Bloch (IReS Strasbourg)
JLIP-ptrel correlation in : b jets: small Et dependence:b = 1.00 in average, = 1.0050.010 . if Et<55GeV, = 0.9800.010 . if 55<Et<85GeV c and light jets:constant valuecl = 0.920.03 D. Bloch (IReS Strasbourg)
double/single tag efficiency ratio are compatible, for any JLIP cut: factor = 1.0370.005 D. Bloch (IReS Strasbourg)
other SystemD factors • Same b-tag efficiency if = 10.3 negligible systematics • Varying the ptrel cut between 0.3 and 1.0 GeV/c: seems to jump between 2 solutions in real data (for JLIP Prob<0.007) take the average = 0.385 0.015 • Overall result: (for JLIP Prob<0.007) = 0.385 0.009 (stat) . 0.010 (b) 0.003 () 0.015 (ptrel) = 0.385 0.009 (stat) 0.018 (syst) D. Bloch (IReS Strasbourg)
Test SystemD statistical error Repeat 100,000 toy MC measurements good agreement with the evaluated error in provided macro D. Bloch (IReS Strasbourg)
mistag systematics • take negative tags as the average of jet trigger and emqcd data: use half of the difference as a systematic: ~ 5-10% relative uncertainty on mistags • Fll = positive/negative light quark tag: - compare QCD, Zuu/dd/ss and W+jets MC - test if the V0 fraction is similar in MC and data • Fhf = all negative/light negative tags in QCDMC: vary the c and b content in QCD MC ! • but Fll and Fhf systematics not yet looked at… D. Bloch (IReS Strasbourg)