Barbara Storaci, Nicola Serra, Niels Tuning Bs →μ+μ- Bfys-meeting, 15 th May 2009 1.

Slides:



Advertisements
Similar presentations
Statistical Issues at LHCb Yuehong Xie University of Edinburgh (on behalf of the LHCb Collaboration) PHYSTAT-LHC Workshop CERN, Geneva June 27-29, 2007.
Advertisements

NASSP Masters 5003F - Computational Astronomy Lecture 5: source detection. Test the null hypothesis (NH). –The NH says: let’s suppose there is no.
0 25. Sept 2006 M.Smizanska, Lancaster University, UK LHC preparations for precise measurements of muonic very rare B-decays 25. Sept 2006 M.Smizanska,
9-11 October 2006Flavour in the era of LHC - F.Teubert1 Search for the decay B s  µ + µ - at LHCb Frederic Teubert CERN, PH Department on behalf of the.
Search for B     with SemiExclusive reconstruction C.Cartaro, G. De Nardo, F. Fabozzi, L. Lista Università & INFN - Sezione di Napoli.
Peter Fauland (for the LHCb collaboration) The sensitivity for the B S - mixing phase  S at LHCb.
B Production and Decay at DØ Brad Abbott University of Oklahoma BEACH 2004 June 28-July 3.
G. Cowan Lectures on Statistical Data Analysis Lecture 10 page 1 Statistical Data Analysis: Lecture 10 1Probability, Bayes’ theorem 2Random variables and.
G. Cowan Lectures on Statistical Data Analysis 1 Statistical Data Analysis: Lecture 7 1Probability, Bayes’ theorem, random variables, pdfs 2Functions of.
G. Cowan RHUL Physics Bayesian Higgs combination page 1 Bayesian Higgs combination using shapes ATLAS Statistics Meeting CERN, 19 December, 2007 Glen Cowan.
Winter Meeting, Santiago de Compostela Bs  μ+μ- in LHCb PROGRAMA NACIONAL DE BECAS FPU Diego Martínez Santos Universidade de Santiago de Compostela (USC)
W properties AT CDF J. E. Garcia INFN Pisa. Outline Corfu Summer Institute Corfu Summer Institute September 10 th 2 1.CDF detector 2.W cross section measurements.
Irakli Chakaberia Final Examination April 28, 2014.
1 Methods of Experimental Particle Physics Alexei Safonov Lecture #23.
M. Adinolfi - University of Bristol1/19 Valencia, 15 December 2008 High precision probes for new physics through CP-violating measurements at LHCb M. Adinolfi.
1 TTU report “Dijet Resonances” Kazim Gumus and Nural Akchurin Texas Tech University Selda Esen and Robert M. Harris Fermilab Jan. 26, 2006.
1 Rare Bottom and Charm Decays at the Tevatron Dmitri Tsybychev (SUNY at Stony Brook) On behalf of CDF and D0 Collaborations Flavor Physics and CP-Violation.
V.Petracek TU Prague, UNI Heidelberg GSI Detection of D +/- hadronic 3-body decays in the CBM experiment ● D +/- K  B. R. 
Paper on J/  and b production Wenbin Qian, Patrick Robbe for the F-WG, Tsinghua Beijing/LAL Orsay, 2 Dec 2009.
36th ITEP Winter School of Physics (2008) 1 Bs  μ+μ- in LHCb PROGRAMA NACIONAL DE BECAS FPU Diego Martínez Santos (ДИЕГО МАРТИНЕЗ САНТОC) Universidade.
G. Cowan Lectures on Statistical Data Analysis Lecture 8 page 1 Statistical Data Analysis: Lecture 8 1Probability, Bayes’ theorem 2Random variables and.
1 Marek Szczekowski, Artur Ukleja Warsaw Group 3 June 2015 Model-independent search for CPV in three-bodies charm baryon decays Outline Selections for.
Lecture 8 Source detection NASSP Masters 5003S - Computational Astronomy
A bin-free Extended Maximum Likelihood Fit + Feldman-Cousins error analysis Peter Litchfield  A bin free Extended Maximum Likelihood method of fitting.
DIJET STATUS Kazim Gumus 30 Aug Our signal is spread over many bins, and the background varies widely over the bins. If we were to simply sum up.
ES 07 These slides can be found at optimized for Windows)
1 Warsaw Group May 2015 Search for CPV in three-bodies charm baryon decays Outline Selections Mass distributions and reconstructed numbers of candidates.
Moritz Backes Measuring Correlations. Motivation e.g.: To use two-dimensional sidebands in data-driven background estimation techniques the control samples.
Barbara Storaci, Wouter Hulsbergen, Nicola Serra, Niels Tuning 1.
Jeroen van Hunen (for the LHCb collaboration) The sensitivity to  s and  Γ s at LHCb.
06/2006I.Larin PrimEx Collaboration meeting  0 analysis.
1 The use of control channels in the analysis of the B s  μ + μ - decay Tuesday Meeting, 2 June 2009.
Charm Mixing and D Dalitz analysis at BESIII SUN Shengsen Institute of High Energy Physics, Beijing (for BESIII Collaboration) 37 th International Conference.
1 D *+ production Alexandr Kozlinskiy Thomas Bauer Vanya Belyaev
Investigation on CDF Top Physics Group Ye Li Graduate Student UW - Madison.
23 Jan 2012 Background shape estimates using sidebands Paul Dauncey G. Davies, D. Futyan, J. Hays, M. Jarvis, M. Kenzie, C. Seez, J. Virdee, N. Wardle.
Using RooFit/RooStat in rare decay searches Serra, Storaci, Tuning Rare Decays WG.
1 Rare B Decays At CDF Michael Weinberger (Texas A&M University ) For the CDF Collaboration DPF 2006 November 1, 2006.
(Day 3).
Status of the Higgs to tau tau
Statistical Significance & Its Systematic Uncertainties
Theoretical Motivation Analysis Procedure Systematics Results
Muon momentum scale calibration with J/y peak
Recent B physics results from BABAR and BELLE
Heavy Flavor Results from CMS
Multi-dimensional likelihood
Bs  μ+μ- in LHCb Diego Martínez Santos
An Important thing to know.
° status report analysis details: overview; “where we are”; plans: before finalizing result.. I.Larin 02/13/2009.
W boson helicity measurement
B  at B-factories Guglielmo De Nardo Universita’ and INFN Napoli
Bs  μ+μ- in LHCb Diego Martínez Santos
CP violation in Bs→ff and Bs → K*0K*0
Computing and Statistical Data Analysis / Stat 7
B Physics at the LHC Neville Harnew University of Oxford.
Study of e+e collisions with a hard initial state photon at BaBar
Application of CLs Method to ATLAS Higgs Searches
LHCb Rare Decays Status
LHCb Rare Decays Status
Paul Sail, Lars Eklund and Alison Bates
CHAPTER – 1.2 UNCERTAINTIES IN MEASUREMENTS.
Dilepton Mass. Progress report.
Search for Lepton Flavour Violation in the decay  → BaBar
° status report analysis details: overview; “where we are”; plans: before finalizing result.. I.Larin 02/13/2009.
Paul Sail, Lars Eklund and Alison Bates
Paul Sail, Lars Eklund and Alison Bates
CHAPTER – 1.2 UNCERTAINTIES IN MEASUREMENTS.
Uncertainty Propagation
Presentation transcript:

Barbara Storaci, Nicola Serra, Niels Tuning Bs →μ+μ- Bfys-meeting, 15 th May

Index Bs → μ + μ - analysis strategy Goal Toy Model description Comparison of analysis methods 1-bin approach Modified Frequentist Approach (MFA) Unbinned Approach Results and Conclusions 2 Talk given at the last Bs → μ + μ – working group meeting (7 th May 2009) :

Bs → μ + μ - 3 FCNC process: Sensitive to new physics Rare Decay [ SM: BR = (3.35 ± 0.32) ]: Precise prediction: small hadronic uncertainties thanks to ΔMs measurement. In all MSSMs BR(Bs → μ + μ - ) prop to tan 6 β sensitive to models with high tanβ In NUHM (which contains cMSSM) prediction of BR(Bs → μ + μ - ) ~ Beyond Tevatron discovery possibility, but early measurement for LHCb One of the key channels for (early) NP discovery

Analysis Strategy (1) Pre-Selection PV   2 <14 IPS <6 Invariant Mass Likelihood (IML)[± 600MeV] Z pos SV downstream respect to PV 4 Selection Invariant Mass Likelihood (IML) [± 60MeV] Particle Identification Likelihood (PIDL) Geometrical Likelihood (GL) > 0.5 Variables built and chosen to be uncorrelated (see roadmap!) BR extraction/ CL limit 3 possible methods 1-bin Approach Modified Frequentist Approach Unbinned Approach TOPIC of this TALK Only for completeness: GL built with:  (Bs) IPS (Bs) DOCA (distance of closest approach between the 2 μ) Isolation of the 2 μ

Analysis Strategy (2) Pre-Selection PV   2 <14 IPS <6 Invariant Mass Likelihood (IML)[± 600MeV] Z pos SV downstream respect to PV 5 Selection Invariant Mass Likelihood (IML) [± 60MeV] Particle Identification Likelihood (PIDL) Geometrical Likelihood (GL) > 0.5 Variables built and chosen to be uncorrelated BR extraction (CL limit) 3 possible methods 1-bin Approach Modified Frequentist Approach Unbinned Approach TOPIC of this TALK Only for completeness: GL built with:  (Bs) IPS (Bs) DOCA (distance of closest approach between the 2 μ) Isolation of the 2 μ

Goals Extraction of BR with different methods Study for different signal (BR) Assumptions: Input: generated with toy model All methods used same toy experiments Complete factorization of GL and Bs Mass Used large sample to describe signal (equivalent to control channel in real data) Side-Band perfectly describes peak region 6

Toy Model Description (1) Only tw0 variables: GL, Bs Mass Signal: Polynomial Background: Exponential e x Signal: Gaussian Background: Polynomial Exp to match roadmap prediction in the sensitive region (GL>0.5) Bg = GL(exp)  Bs(pol) Sg = GL(pol)  Bs(gauss) Peak_pdf = Nb [GL(exp)  Bs(pol)] + Ns [GL(pol)  Bs(gauss)] 7

Toy Model Description (2) For each toy, we produce three statistically independent samples: E.g. 1 year, BR= → “ Control Channel” → “ Side Band” (e.g. [M Bs -120, M Bs -60]  [M Bs +60, M Bs +12o]) → “ Peak Region” Extract BR (Ns) with different methods 8 N.B. all numbers Poison fluctuating

Analysis Methods 1-bin Approach (Simplest): Rectangular cuts on variables to select a bin (mass- window, cut on GL) Modified Frequentist Approach (Roadmap): 3 bins, GL[0.5 – 1.0], 5 bins, Bs Mass [ – ] Estimation of probability per each bin (for an assumed BR value) of the hypotheses signal + background or only background Un-binned Approach (This talk): “Generalization” of MFA in an unbinned way (see later) 9

1-bin Approach Sensitivie region! GL Bs Mass Mass window GL cut at 0.76 GL cut optimization with Figure Of Merit (FOM) for a 3 sigma sensitivity: Real BR = (red line) 3 years of data taken N toys = 500 → The method is unbiased Measurement / Sensitivity /

MFA method Box of two variables: Bs mass, GL 3 bins, GL[0.5 – 1.0] 5 bins, Bs Mass [ – ] For given BR, estimate probability that data in a bin are signal+background or only background Calculate: GL Bs Mass Sensitive region! 11

Br extraction with MFA Calculating ΔΧ 2 for different BR hypotheses Probability of a certain amount of signal (assuming a Br) hypothesis with the expected background (from side-band) Probability of having only background Minimum from parabola interpolation → The method is unbiased Real BR = (red line) 3 years of data taken N toys = 500 Measurement / Sensitivity / Added by us!!!

Why an unbinned strategy? No bin size optimization necessary Gain sensitivity to BR? Used a flexible tool (RooFit) → easy to extend for more dimensions and more background sources 13

Unbinned method Fit: Background (i.e. Side-Band) (e.g. [M Bs -120, M Bs -60]  [M Bs +60, M Bs +12o]) pdf_bg’ = GL(RooKeysPdf)  Bs(Pol) (Extended pdf ) Signal (i.e. Control-Channel) pdf_sg’ = GL(RooKeysPdf)  Bs(gauss) (Extended pdf) Use the result to build: pdf_tot’ = Nb pdf_bg’ + Ns pdf_sg’ Fit data with pdf_tot’ to obtain Ns (for the moment parameter Nb fitted from Side Band, work in progress for a simultaneous fit) GL Bs Mass Sensitive region! Use NO knowledge from generation:  GL: no assumption on signal and background shape  Bs Mass: reasonable assumption 14

RooKeysPdf: idea Technique to produce a continuous estimates of f(x) of the parent distribution without assume a model (non- parametric) (Direct generalization of the Averaged Shifted Histograms (ASH) techniques) Convolution of Gaussian with different sigmas drawn under each point of the dataset Kernel Estimation in High-Energy Physics (auth: Cranmer K.) ONLY for illustrative purpose!!!  Accurate estimation of parent distribution with narrow kernels in high statistics region  Wide kernels to smooth out statistical fluctuations in low-density regions 15

Preliminary Results: SM 2 years1 year 3 years MFA and unbinned methods unbiased Unbinned method seems to give better RMS Both methods run on the SAME toys!!! 16

Preliminary Results: Br= year 1 year 0.5 year Both methods run on the SAME toys!!! Both methods unbiased Unbinned method seems to be good also at low statistics 17

3 methods comparison 3 years In ideal conditions all methods work well (Remember: cuts were optimized in 1-bin method) MFA lies in between the 1-bin and unbinned methods (no bin size optimization for MFA here) 18

Conclusions Implementation of BR extraction in the Modified Frequentist Approach First investigation of an alternative method (unbinned) to extract BR in the Bs → μ + μ - All methods unbiased in ideal conditions Unbinned approach good in extreme conditions: small BR, small amount of statistics (early measurement) 19