CMS DATA ANALYSIS SCHOOL |
Jan 25-29 2011 at LPC, FNAL

From Benchmarks of the Standard Model to First Discoveries

Registration for the School and the agenda are at:
http://indico.cern.ch/conferenceDisplay.py?confId=112319
(Please note: to register, a CERN e-mail (NICE) account is required.)

Course Area at CERN for the CMS Data Analysis School Jan 2011
https://espace.cern.ch/learncms/AnalysisSchoolJan11/default.aspx

Speakers for opening session

  • JoAnne Hewett (SLAC) - The Big Picture
  • Eric Prebys (FNAL) - The LHC
  • Joel Butler (FNAL) - The CMS Detector
  • Gigi Rolandi (CERN) - CMS Results and the Next Two Years

EJTERM Group Photo

Group Photo View Full Size

This school was formerly known as EJTERM. A link to the EJTERM 2010 site can be found here.

All students are required to complete 4 sets of pre-workshop exercises. They can be accessed via the link above for the Course Area. On the left menu of the Course Area, click on "Pre-exercises" and then access them. To answer questions in the "Pre-exercises," you must click "Submit Your Pre-exercises" on the left menu in the Course Area.

Please follow the link below for a list of local restaurants, map of Fermilab, etc.

http://www-ppd.fnal.gov/conf-w/UsefulLinks.htm

Access to Fermilab for CMSDAS participants does not require a Fermi ID. Simply tell the guard at the entrance that you are attending CMSDAS.


Exercises

At the School, there will be two types of exercise...short: ~2 hrs, long ~ 2 days

Schedule

Day 1 & 2 + morning Day 3 short exercises interspersed with talks
Day 3 afternoon, Day 4, Day 5 morning long exercises.

An agenda (under development) is here: http://indico.cern.ch/conferenceDisplay.py?confId=112319


The exercises for the School will be available just prior to the workshop. Descriptions will be made available somewhat earlier. They will contain many topics geared toward discovery physics. The complete set of exercises from EJTERM 2010 are available at the EJTERM 2010 website.

The Pre-CMSDAS Exercises have been developed by


Sudhir Malik
University of
Nebraska-Lincoln

Chris Jones
Cornell University

Charles Plager
UCLA

Eric Vaandering
Fermilab

The Pre-CMSDAS Exercises and CMSDAS Exercises have been tested by:


Yu Zheng
Purdue University

Zhen Hu
Purdue University

Fan Yang
Fermilab

Eric Brownson
Vanderbilt University

Cesar Pollack
University of
Puerto Rico - Mayaqüez

Alexey Svyatkovskiy
Purdue University

Short Exercises - Duration ~ 2 hrs

1) Jets


Kalanand Mishra
Fermilab

Rob Harris
Fermilab

Eva Halkiadakis
Rutgers University

John Paul Chou
Brown University

Jason St. John
Boston University

Ilya Osipenkov
Texas A&M University

Chiyoung Jeong
Texas Tech University

Description: The short exercise provides hands-on experience accessing jets, plotting basic jet quantities, and applying jet energy corrections and their uncertainty. You will become familiar with basic jet types and algorithms and how to use them in your analysis.

Link:
https://twiki.cern.ch/twiki/bin/view/CMS/SWGuideCMSDataAnalysisSchoolJetAnalysis

2) Generators (sequence of 3)


Stephen Mrenna

Charles Plager

Description: These tutorials will guide you on an investigation of the similarities and differences that can be encountered when comparing predictions made with different Monte Carlo (MC) tools. In all cases, we focus on W+ and W- production and decay to leptons at the LHC (a 7 TeV, proton-proton collider). We are using a ROOT-based analysis package similar to the one used by the Generator Services (GENSER) group for release-to-release validation of MC tools distributed by CERN. Since this package makes use of the HepMC standard for MC event records, it is called /HepMCAnalysis/. We rely on the WplusJets analysis tool.

Link:
https://twiki.cern.ch/twiki/bin/view/CMS/SWGuideCMSDataAnalysisSchoolGeneratorExercise

3) ROOSTATS (sequence of 3)


Petar Maksimovic

Gena Kukartsev

Bob Cousins

Jordan Tucker

Description: ExoStive RooSting

The exercise introduces methods and tools for answering very common statistics questions while complying with rigorous standards set in today's experimental particle physics. The main topic is the ways to find a confidence interval for a parameter of interest. We cover the most popular tecniques: profile likelihood, bayesian (including MCMC), Feldman-Cousins. Participants will learn how to use RooStats as the software framework for statistics, and RooFit as the language for data modeling. We overview the popular cases of a counting experiment, a shape analysis and a combination of multiple analysis channels. We also discuss how to include systematic uncertainties and prior information into the interval calculations. In the beginning of the exercise, we offer a theoretical overview of the basic concepts.

The exercise will consist of three 2-hour sessions and will be run twice during the school. However, the two runs will not be identical: the first exercise run will start with two hours of the overview lecture, followed by the second hands-on section dedicated to RooFit and the final practical section on RooStats. We only offer the lecture once, so the second exercise run will start with the section on RooFit, followed by two RooStats sections.

Link:
https://twiki.cern.ch/twiki/bin/view/CMS/SWGuideCMSDataAnalysisSchoolExoStiveRooStingExercise

4) Btag & Vertexing


Francisco Yumiceva

Meenakshi Narain
Brown University

Description: Learn about the properties and differences between primary vertices and the luminous region, also known as beam spot. Students will learn how to access the vertex collections and make plots of the main variables.

Learn how to identify b-flavored jets and plot the b-tag efficiency and mistagging rate as a function of jet kinematics for a given algorithm. Use the b-tagging scale factors derived from data to correct the Monte Carlo efficiencies.

Link: https://twiki.cern.ch/twiki/bin/view/CMS/SWGuideCMSDataAnalysisSchoolbTagExercise

5) Tracking


Kevin Burkett
Fermilab

Jim Pivarski

Description:
The goals of the tracking exercise are:

  • learn how to access the collections of tracks and vertices in the event
  • learn how to search for tracking variables relevant for your analysis ("shopping for data")
  • learn how to work with the track geometry: calculating a vertex
  • learn to build invariant mass distributions and search for resonances.

Link:
https://twiki.cern.ch/twiki/bin/view/CMS/SWGuideCMSDataAnalysisSchoolIntroductoryTrackingExercise

6) Photons


Yuri Gershtein
Rutgers University

Andrew Askew
The Florida State University

Marat Gataullin

Description: The three short photon exercises are designed to cover the breadth of photon identification for beginners. Three basic elements of photon selection (discriminating against jets) and analysis are covered: the characteristic shower in the ECAL, the isolation in the immediate vicinity, and the measurement of efficiency (with emphasis on trigger).

Exercise 1: Shower Shape
Exercise 2: Isolation
Exercise 3: Measurement of HLT efficiency

Link:
https://twiki.cern.ch/twiki/bin/view/CMS/SWGuideCMSDataAnalysisSchoolPhotonShortExercise

7) Electrons


Jeff Berryhill

Bryan Dahmes

Paolo Meridiani

Description: Detecting high energy electrons in CMS data, at the electroweak scale and beyond.

The exercise is aimed at describing a selection of clean isolated electrons (from W and Z), separating them from the background from QCD fakes, non isolated electrons and electrons from conversion. We will use an inclusive sample of triggered electrons and describe the selection variables, how to check them, and see step by step how the W and Z signals emerge from the background. At the end of the exercise, the successful student will be comfortable with arriving at their own electron selection and estimating its performance on data.

Link:
https://twiki.cern.ch/twiki/bin/view/CMS/SWGuideCMSDataAnalysisSchoolElectronShortExercise

8) Particle Flow


Rick Cavanaugh

Lucie Gauthier

Chris Silkworth

Vivian O'Dell

Description: Particle-flow algorithms aim to individually reconstruct and identify all stable particles in the event, (electrons, muons, photons, charged hadrons and neutral hadrons) by comprehensively using all sub-detectors in an optimal way to determine each particle's direction, energy and type. The list of individual reconstructed particles in the event can then be used as the starting point to analyze an event, as if the list had come from a Monte-Carlo event generator. For example, using reconstructed particles, one can build jets, determine the missing transverse energy, reconstruct and identify taus from their decay products, quantify charged lepton isolation with respect to other particles, help tag b-jets, etc.

This short exercise briefly motivates the basic conditions needed for a successful particle-flow algorithm. Each participant will build a simplistic, toy particle-flow algorithm from basic sub-detector elements and use that algorithm to identify and reconstruct their own list of particles in very simple events. Participants will then compare results from their simple, toy algorithm with traditional calorimeter-only techniques and with the actual CMS particle-flow algorithm itself. The goal is to enable participants to learn the essential ingredients needed for any particle-flow algorithm and, by the end of the exercise, gain a basic understanding of particle-flow reconstruction techniques as well as an appreciation for the advantages of performing one's analysis at the particle-level. At the end of the exercise, participants will be able to place their simple, toy algorithm in the context of the sophisticated, real CMS particle-flow algorithm, which will be briefly summarised with an emphasis on the many interesting and intricate cases which the real algorithm handles.

Link:
https://twiki.cern.ch/twiki/bin/view/CMS/SWGuideCMSDataAnalysisSchoolPFlowShortExercise

9) Muons


Adam Everett
Purdue University

Jordan Tucker

Description: The short exercise provides hands-on experience accessing muons, plotting basic muon quantities (kinematic, fit, quality), and applying different muon object and event selection cuts to increase sample purity. You will become familiar with basic muon types and specialized muon fit algorithms and how to use them in your analysis.

Link:
https://twiki.cern.ch/twiki/bin/view/CMS/SWGuideCMSDataAnalysisSchoolMuonExercise

10) Event Visualization


Dan Green
Fermilab

Sudhir Malik
University of Nebraska-Lincoln

Description: Being able to examine an individual event in detail is an important tool for our physics analyses. Searches will always operate on the extreme edges of phase space where a handful of events are of primary importance.

This short exercise uses the early 2010 data from CMS with data extracted from the "Exotica Scan". The users should download and make sure Fireworks functions on their laptop prior to the exercise so as to avoid startup delays.

First, navigation and features of Fireworks are explored - commands, buttons, views and collections. Then events are scanned; jets from April 2010, soft muon from b decay, Missing ET ( "core" and tails, ECAL and HCAL spikes, fake muons, and cracks), photons, W and Z.

Finally, top events appeared just before ICHEP as the luminosity appeared. Kinematic reconstruction of a top pair should be attempted, as was, in fact, done to fully convince ourselves that we were observing top events.

Link:
https://twiki.cern.ch/twiki/bin/view/CMS/SWGuideCMSDataAnalysisSchoolEventScanningExercise

11) MET and High-MET Event Scan


Dinko Ferencek

Jordan Damgov

Description: Missing ET is one of the most important observables used in searches for new physics beyond the Standard Model. The first part of this exercise will introduce CMS newcomers to different MET reconstruction algorithms available in CMS and will provide hands-on experience accessing different types of MET objects using CMSSW (both within and outside the PAT framework). The second part will be devoted to a scan of high-MET events from the 2010 collision data where events are selected and inspected using the existing CMSSW tools.

Link:
https://twiki.cern.ch/twiki/bin/view/CMS/SWGuideCMSDataAnalysisSchoolMETExercise

12) SUSY OSETs


Joe Incandela

Charles Plager
UCLA

Stephen Mrenna

Dinko Ferencek

Description: Imagine you have generated a Monte Carlo sample, but some aspect of your signal is not as desired. For example, you simply made a mistake or had to make a guess about some parameters. This exercise focuses on how to modify your events to improve your signal modeling. In this exercise, you will learn how to access Monte Carlo truth information, calculate probabilities based on them, and throw random numbers to reweight your events. There are three levels to this exercise:

  1. Reweight a top => b H+ sample to t=> b W+
  2. Add spin correlations to a standard t t~ sample
  3. Make an OSET simulation with 3-body decays look like an mSugra model

Link:
https://twiki.cern.ch/twiki/bin/view/CMS/SWGuideCMSDataAnalysisSchoolOSETExercise


Long exercises - Duration ~ 2 days

1) Search for Dijet Resonances


Rob Harris
Fermilab

Eva Halkiadakis
Rutgers University

Kalanand Mishra
Fermilab

John Paul Chou
Brown University

Jason St. John
Boston University

Description: The long exercise will teach you how to search for dijet resonances. You will learn the signal, the QCD background, and the published CMS techniques of searching for new physics with the dijet mass distribution and the dijet ratio. You will search for resonance signals in actual CMS data.

Link:
https://twiki.cern.ch/twiki/bin/view/CMS/SWGuideCMSDataAnalysisSchoolJetResonances

2) Search for the Z'


Adam Everett
Purdue University

Jordan Tucker

Description: The long exercise will teach you how to search for high mass dimuon resonances. You will learn the signal, the QCD background, the ttbar background, and the CMS techniques of searching for new physics with the dimuon mass distribution. You will evaluate analysis specific selection cuts, and you will search for resonance signals in actual CMS data.

Link:
https://twiki.cern.ch/twiki/bin/view/CMS/SWGuideCMSDataAnalysisSchoolZPrimeExercise

3) Top Cross Section


Cecilia Gerber

Francisco Yumiceva

Sadia Khalil

Sal Rappoccio

Description: Measure the top pair production cross section in lepton+jets+MET by simultaneously fitting a sample with 3 jets and 4 or more jets. Students will learn about the main background sources as a function of the number of jets, estimation of the QCD multijet background using a data-driven method, and extraction of the top pair signal.

Link:
https://twiki.cern.ch/twiki/bin/view/CMS/SWGuideCMSDataAnalysisSchoolTopExercise

4) Upsilon Cross Section


Jake Anderson
Fermilab

Yu Zheng
Purdue University

Zhen Hu

Nuno Leonardo
Purdue University

Ian Shipsey
Purdue University

Description: On December 26 2010 CMS posted to arXiv, and submitted to PRD, the first measurement of the Upsilon(nS) differential production cross section at sqrt(s)= 7 TeV, based on an integrated luminosity of 3/pb. This exercise reproduces the measurement and extends it to the full data set (40/pb) for the first time, anticipating the results CMS plans to send to the Moriond Conference in March 2011.

The exercise covers MC simulation, kinematics, the definition and calculation of acceptance, the data driven techniques used to measure the trigger and muon identification efficiencies, signal and background pdf design and both un-weighted and weighted fitting.

Systematic uncertainties are a critical part of any analysis. We discuss and explore sources of systematic uncertainty and methods to estimate them both generally and in the context of the Upsilon measurement. We will use some of the dominant systematic uncertainties affecting the Upsilon cross section to illustrate these Topics.

Link:
https://twiki.cern.ch/twiki/bin/view/CMS/SWGuideCMSDataAnalysisSchoolUpsilonExercise

5) Search for the Higgs (high mass)


Alexey Drozdetskiy

Andrey Korytov

Jonatan Piedra
Universidad de Cantabria

Description: If LHC runs according to expectations - 2011 would be a great year, a great year first for limits and then for possible discoveries.

There are several "high mass Higgs" channels, the main contributor into setting limits is: H -> ZZ -> 2l2nu.

In this exercise we will start with an overview of the differences between signal and background processes and corresponding discriminating variables for the channel; we will continue with up to 3 different analysis scenario (robust; optimized; MVA optimized) to set limits on Higgs cross section. And last step - we will look into several systematic uncertainty contributions (like momentum scale for leptons, JetMET uncertainties, etc.): their evaluation and effect on the final result. Given enough time we will overview and discuss the full list of systematic uncertainties and update final results with that in mind.

We will check all the distributions and results with both data and MC.

Link:
https://twiki.cern.ch/twiki/bin/view/CMS/SWGuideCMSDataAnalysisSchoolHiggsSearchExercise

6) Photons: New Physics


Yuri Gershtein
Rutgers University

Andrew Askew
The Florida State University

Marat Gataullin

Description: There is nothing quite like doing analysis with photons to introduce new students to doing analysis with photons. So we're going to do one! The long exercise will kick off an effort for the search for a three photon final state (which arises naturally from new physics with vector-like confinement). It is anticipated that the majority of the work on this brand-new analysis for CMS will be completed at CMSDAS, and that the involved analyzers will stay involved to see the results through to publication.

Link:
https://twiki.cern.ch/twiki/bin/view/CMS/SWGuideCMSDataAnalysisSchoolPhotonLongExercise

7) Exotica: Black Holes


Alexey Ferapontov

Greg Landsberg

Ka Vang Tsang

Description: In this long exercise, you will learn how to search for microscopic Black Holes in real collision data. The exercise will teach you how to estimate the dominant QCD background using a novel data-driven technique. You will optimize the offline selection criteria and use this information to search for unique Black Holes signatures.

Link:
https://twiki.cern.ch/twiki/bin/view/CMS/SWGuideCMSDataAnalysisSchoolBlackHolesLongExercise