PHYSTAT 05P. Catastini Bias-Free Estimation in Multicomponent Maximum Likelihood Fits with Component-Dependent Templates Pierluigi Catastini I.N.F.N. -

Slides:



Advertisements
Similar presentations
Recent Results on Radiative Kaon decays from NA48 and NA48/2. Silvia Goy López (for the NA48 and NA48/2 collaborations) Universitá degli Studi di Torino.
Advertisements

PHYSTAT2007, CERN Geneva, June 2007 ALICE Statistical Wish-List I. Belikov, on behalf of the ALICE collaboration.
Gavril Giurgiu, Carnegie Mellon 1 B s Mixing at CDF Seminar at Fermi National Accelerator Laboratory Gavril Giurgiu Carnegie Mellon University August 16,
14 Sept 2004 D.Dedovich Tau041 Measurement of Tau hadronic branching ratios in DELPHI experiment at LEP Dima Dedovich (Dubna) DELPHI Collaboration E.Phys.J.
Measurement of non BB Decays of Y(4S) to Y(1S)     and Y(2S)     Silvano Tosi Università & INFN Genova.
ECIV 201 Computational Methods for Civil Engineers Richard P. Ray, Ph.D., P.E. Error Analysis.
ICFP 2005, Taiwan Colin Gay, Yale University B Mixing and Lifetimes from CDF Colin Gay, Yale University for the CDF II Collaboration.
Exclusive D s Semileptonic decays using kinematic fitting.
Basic Measurements: What do we want to measure? Prof. Robin D. Erbacher University of California, Davis References: R. Fernow, Introduction to Experimental.
Exclusive D s Semileptonic decays using kinematic fitting.
Role and Place of Statistical Data Analysis and very simple applications Simplified diagram of scientific research When you know the system: Estimation.
Sensitivity of searches for new signals and its optimization
Measurement of the Branching fraction B( B  D* l ) C. Borean, G. Della Ricca G. De Nardo, D. Monorchio M. Rotondo Riunione Gruppo I – Napoli 19 Dicembre.
2.3 More on Solving Linear Equations
TOF (and Global) PID F. Pierella for the TOF-Offline Group INFN & Bologna University PPR Meeting, January 2003.
Xiaoyan LinQuark Matter 2006, Shanghai, Nov , Study B and D Contributions to Non- photonic Electrons via Azimuthal Correlations between Non-
880.P20 Winter 2006 Richard Kass 1 Confidence Intervals and Upper Limits Confidence intervals (CI) are related to confidence limits (CL). To calculate.
880.P20 Winter 2006 Richard Kass 1 Maximum Likelihood Method (MLM) Does this procedure make sense? The MLM answers this question and provides a method.
A Reconstruction Algorithm for a RICH detector for CLAS12 Ahmed El Alaoui RICH Workchop, Jefferson Lab, newport News, VA November th 2011.
GlueX Particle Identification Ryan Mitchell Indiana University Detector Review, October 2004.
R. Kass/W03P416/Lecture 7 1 Lecture 7 Some Advanced Topics using Propagation of Errors and Least Squares Fitting Error on the mean (review from Lecture.
Max Baak1 Impact of Tag-side Interference on Measurement of sin(2  +  ) with Fully Reconstructed B 0  D (*)  Decays Max Baak NIKHEF, Amsterdam For.
Optimizing DHCAL single particle energy resolution Lei Xia Argonne National Laboratory 1 LCWS 2013, Tokyo, Japan November , 2013.
Irakli Chakaberia Final Examination April 28, 2014.
Non-identified Two Particle Correlations from Run I Understanding drift chamber tracking – Tracker and candidatory – Two particle efficiencies/ghosts A.
Optimizing DHCAL single particle energy resolution Lei Xia 1 CALICE Meeting LAPP, Annecy, France September 9 – 11, 2013.
D 0 Measurement in Cu+Cu Collisions at √s=200GeV at STAR using the Silicon Inner Tracker (SVT+SSD) Sarah LaPointe Wayne State University For the STAR Collaboration.
BNL/ Tatsuya CHUJO CNS workshop, Tokyo Univ. Identified Charged Single Particle Spectra at RHIC-PHENIX Tatsuya Chujo (BNL) for the PHENIX.
SUPA Advanced Data Analysis Course, Jan 6th – 7th 2009 Advanced Data Analysis for the Physical Sciences Dr Martin Hendry Dept of Physics and Astronomy.
Calibration of the new Particle Identification Detector (PID) Tom Jude, Derek Glazier, Dan Watts.
Chapter 2 Section 3 Copyright © 2008 Pearson Education, Inc. Publishing as Pearson Addison-Wesley.
A statistical test for point source searches - Aart Heijboer - AWG - Cern june 2002 A statistical test for point source searches Aart Heijboer contents:
Gavril Giurgiu, Carnegie Mellon, FCP Nashville B s Mixing at CDF Frontiers in Contemporary Physics Nashville, May Gavril Giurgiu – for CDF.
Section 10.1 Confidence Intervals
Lecture 4: Statistics Review II Date: 9/5/02  Hypothesis tests: power  Estimation: likelihood, moment estimation, least square  Statistical properties.
PROBABILITY AND STATISTICS FOR ENGINEERING Hossein Sameti Department of Computer Engineering Sharif University of Technology Principles of Parameter Estimation.
Measurement of Vus. Recent NA48 results on semileptonic and rare Kaon decays Leandar Litov, CERN On behalf of the NA48 Collaboration.
OPERA Neutrino Experiment Tija Sīle presentation is based on: Doktorantūras skolas “Atomāro un nepārtrauktās vides fizikālo.
1 Methods of Experimental Particle Physics Alexei Safonov Lecture #15.
PID simulations Rikard Sandström University of Geneva MICE collaboration meeting RAL.
Geant4 Tracking Test (D. Lunesu)1 Daniela Lunesu, Stefano Magni Dario Menasce INFN Milano GEANT4 TRACING TESTs.
1 Introduction to Statistics − Day 4 Glen Cowan Lecture 1 Probability Random variables, probability densities, etc. Lecture 2 Brief catalogue of probability.
I.BelikovCHEP 2004, Interlaken, 30 Sep Bayesian Approach for Combined Particle Identification in ALICE Experiment at LHC. I.Belikov, P.Hristov, M.Ivanov,
PID Detector Requirements for Emittance Measurement Chris Rogers, MICE PID Review, Thursday Oct 12.
1 Introduction to Statistics − Day 3 Glen Cowan Lecture 1 Probability Random variables, probability densities, etc. Brief catalogue of probability densities.
Particle Identification. Particle identification: an important task for nuclear and particle physics Usually it requires the combination of informations.
Study of b quark contributions to non-photonic electron yields by azimuthal angular correlations between non-photonic electrons and hadrons Shingo Sakai.
Feasibility study of Heavy Flavor tagging with charged kaons in Au-Au Collisions at √s=200 GeV triggered by High Transverse Momentum Electrons. E.Kistenev,
Radia Sia Syracuse Univ. 1 RICH 2004 Outline:  The CLEO-III RICH Detector  Physics Requirements  CLEO-III RICH at work… Performance of the CLEO-III.
1 Methods of Experimental Particle Physics Alexei Safonov Lecture #9.
October 2011 David Toback, Texas A&M University Research Topics Seminar1 David Toback Texas A&M University For the CDF Collaboration CIPANP, June 2012.
Kalanand Mishra June 29, Branching Ratio Measurements of Decays D 0  π - π + π 0, D 0  K - K + π 0 Relative to D 0  K - π + π 0 Giampiero Mancinelli,
G. Cowan Lectures on Statistical Data Analysis Lecture 5 page 1 Statistical Data Analysis: Lecture 5 1Probability, Bayes’ theorem 2Random variables and.
Particle identification by energy loss measurement in the NA61 (SHINE) experiment Magdalena Posiadala University of Warsaw.
Computational Intelligence: Methods and Applications Lecture 26 Density estimation, Expectation Maximization. Włodzisław Duch Dept. of Informatics, UMK.
Proton to Pion ratio in Jet and Bulk region in Heavy Ion collisions Misha Veldhoen (Utrecht University) For the ALICE collaboration Hard Probes 2012 Cagliari,
Simulation of Heavy Hypernuclear Lifetime Measurement For E Zhihong Ye Hampton University HKS/HES, Hall C Outline: 1,Physics 2,Detectors 3,Events.
Measurement of the Muon Charge Ratio in Cosmic Ray Events with the CMS Experiment at the LHC S. Marcellini, INFN Bologna – Italy on behalf of the CMS collaboration.
I'm concerned that the OS requirement for the signal is inefficient as the charge of the TeV scale leptons can be easily mis-assigned. As a result we do.
Mesons in ep missing spectra G. M. Urciuoli. Meson photoproduction provide a good tool to study nucleon resonances. For example, a large number of resonances.
TOF detector in PHENIX experiment PHENIX time-of-flight counter The PHENIX time-of-flight (TOF) counter serves as a particle identification device for.
Copyright © Cengage Learning. All rights reserved.
Status Report Fenfen An
MICE Collaboration Meeting
Chapter 2 Section 3.
For the BaBar Collaboration
Study of dE/dx Performance in TPC at CEPC
Chapter 2 Section 3.
Identified Charged Hadron Production
Copyright © 2008 Pearson Education, Inc
Presentation transcript:

PHYSTAT 05P. Catastini Bias-Free Estimation in Multicomponent Maximum Likelihood Fits with Component-Dependent Templates Pierluigi Catastini I.N.F.N. - Pisa and Siena University Giovanni Punzi S.N.S. and I.N.F.N - Pisa

PHYSTAT 05P. Catastini Problem Suppose we have a sample of particles generated by a certain physics process produced by our experiment. Suppose we know that the sample is a mixture of different particle types, for example, Pions, Protons and Kaons, but the proportion of each particle type is completely unknown. Of course, our experiment is also equipped with some kind of Particle IDentification (PID) device, providing the measurement of some quantity related to the particle type. We want to measure the fractions of each particle type : f , f p, f k.

PHYSTAT 05P. Catastini A “Real Life” Problem * … * At least for a Physicist… Measuring the particle type fractions is common in particle physics: e.g. understanding the particle produced during the fragmentation of the B mesons (flavor tagging), separating different particle decays... Usually PID information provided by energy loss of charged particle in gas (dE/dx), measurement of Time of Flight, Cherenkov light… Solution obtained performing an unbinned Maximum Likelihood Fit. But remember… The mean of the PID observable strongly depends on particle momentum (which is an additional observable, known event-by- event): Component Dependent Templates ! Electrons Protons Muons Kaons Pions

PHYSTAT 05P. Catastini Please write the Likelihood ! Unfortunately, the Likelihood is not simply:  i ( f i P(pid i | , Mom i ) (WRONG!) Using the above, you may get strongly biased results if the additional observables have different distributions [1]. The reason for the failure is, quoting from [1]: “Whenever the templates used in a multi-component fit depend on additional observables, one should always use the correct, complete Likelihood expression, including the explicit distributions of all observables for all classes of events“ In our problem, the above means that we need to include the momentum distributions of each particle type (they are almost always different in practice). [1] physics/ (G.Punzi, PHYSTAT03)

PHYSTAT 05P. Catastini Writing the Likelihood… Particle IDentification information is represented by a certain observable called pid; we than write the likelihood as: Given: f  + f P + f K = 1 j = Pion, Proton, Kaon L (f  f P f K ) =  i ( f  P(pid i, Mom i |  ) + f P P(pid i, Mom i | P) + (1 - f  - f P )P(pid i, Mom i | K) ) =  i (  j f j P(pid i | Mom i, type j ) P(Mom i | type j ) )

PHYSTAT 05P. Catastini A toy study of the “Real Life” Problem f  = 0.50 f P = 0.15 f K = 0.35   = 1.00 GeV/c  P = 1.25 GeV/c  K = 1.50 GeV/c   = 0.25 GeV  P = 0.25 GeV  K = 0.25 GeV Momentum (GeV/c) We generate a sample with known particle types composition as follow: PID variable is distributed according a typical resolution function (i.e. the template used in the fit) defined as PID mes - PID exp (mom): Momenta are distributed according a gaussian N( ,  ) : P(pid i | Mom i, type j ) P(Mom i | type j )

PHYSTAT 05P. Catastini Result of the Fits PionsProtons OK ! If we wouldn’t take into account the momentum distributions… PionsProtons Bias !

PHYSTAT 05P. Catastini Often in “Real Life”… Writing the complete likelihood with all observables distribution is almost straightforward. Of course, provided the assumption you can easily obtain a parameterization of those distribution… Often we have poor information about those distribution (barely acceptable, after a very hard work!), sometimes they could be even completely unknown. If, for example, the goal of the particle type fit we have been performing in the previous slides is to estimate the fractions of particle produced during the heavy quarks fragmentation… Grate! We have no idea about the functional form of each particle type’s momentum distribution. How can we write the correct Likelihood ?

PHYSTAT 05P. Catastini A solution No functional form is known in order to parameterize the missing P(Mom | type). Use a general functional form Series Expansion P(Mom | type j ) =  m a mj F m (Mom) with a mj free parameters of the fit We decide to use Orthogonal Polynomials, among them: Legendre Polynomials P i [-1,1] First type Chebyshev Polynomials T i [-1,1] Second type “ “ U i [-1,1] Lagerre Polynomials L i [0,+  ] Hermite Polynomials H i [- ,+  ] Used from 0 th to 6 th term.

PHYSTAT 05P. Catastini Our toy Replacing the exact distribution N( ,  ) with  m a mj F m (Mom) for each particle type, we fitted again our toy sample: The Bias is again corrected ! Pions OK ! Protons OK !

PHYSTAT 05P. Catastini Some Comments Of course, we are happy: although we didn’t know a priori the P(Mom | type) we have been able to avoid the bias. Please, notice that resolution on the parameter is not degraded a lot ! Just 7 terms of the series expansion were used! Not so many. Projections of P(Mom | type j ) =  m a mj F m (Mom) : Pions Protons Kaons Momentum (GeV/c)

PHYSTAT 05P. Catastini Another Complication Suppose our PID information is obtained by the measurement of the Time Of Flight (TOF). The expression of the Expected TOF is a function of 2 obsevables : TOF exp = arclength / c  sqrt(1 + m j 2 /Mom 2 ) It means that our pdf is ( after having verified that the correlation between arclength and momentum is almost zero ) : P(Mom, Arcl, pid |type j, )  P(pid | Mom, Arc, type j ) * P(Mom | type j ) * P(Arc | type j ) Both unknown ! We want to apply the same technique of series expansion both for momentum and arclength !

PHYSTAT 05P. Catastini Back to our toy L (f j, a mj, b lj ) =  i (  j f j ( P(pid i | Mom i, Arc i, type j )  m a mj F m (Mom i )  l b lj F l (Arc i )) ) Again we used 7 terms for the momenta series expantion We used 3 terms for the arclegth series expantion Fractions, pid and momentum variables generated as before Arclength distributed according a gaussian N( ,  )   =  K =  P   =  K =  P Same distribution for all particle types but in principle you don’t know !

PHYSTAT 05P. Catastini Results Pions OK ! Protons OK ! Projections of P(Arc | type j ) =  l b lj F l (Arc) : Pions Protons Kaons

PHYSTAT 05P. Catastini Conclusions We faced a common problem of particle physics were the incomplete Likelihood expression is cause of a detectable bias. We had cure it ! The proposed problem has also the complication of the lack of information about the distribution of an observable! We solved the problem, removing the bias in the fit results, including series expansion as a parameterization of the unknown distributions (coefficients free parameters determined by the fit). We even faced the case where two observables have unknown distributions. Again we used two different series expansions in order to parameterize those distribution and avoid the bias.