Electron Tag Probe Jeffrey Berryhill Kalanand Mishra Fermilab
Electron Tag & Probe Jeffrey Berryhill Kalanand Mishra Fermilab [ work with muon POG: Nadia Adam, Adam Everett, Adam Hunt & electron POG: Jason Haupt ] CSA 08 Electron May 11, 2008
Introduction to tag-and-probe method - The tag-and-probe method is a powerful tool to measure reconstruction efficiency from data. - Desirable to have a common interface and machinery for “data-driven” electron reconstruction / identification efficiency (and fake rate) in CMS. - We are working with Muon POG people to have the same basic framework for both muon and electron tag-and-probe software. Concept • Utilize data from a clean dielectron Tag-probe invariant mass Z CSA 08 Electron May 11, 2008 sample, e. g. , Z e+e–. • A "tag" is a well identified electron, e. g. , isolated Gsf. Electron in barrel. • The existance of a tag electron usually indicates the existence of the other leg from the same Z. • A "probe" is the other electron coming from the Z. • The method begins to count when both "tag" and "probe" exist. Kalanand Mishra, Fermilab 2
How it works: Algorithms & Design n n The code in the analyzer and producers has been designed to be as generic as possible. The tag-probe maps have therefore been set up with a global efficiency measurement composed of many smaller pieces in mind. For many of the "sub-efficiency" measurements, e. g. , reconstruction efficiency, identification efficiency, L 1 / HLT efficiency, … etc, the efficiency is measured by taking some set of probes that compose the denominator, and finding the sub-set of these probes that pass the desired efficiency criteria as the numerator (the "passing probes"). Extremely important to compare “apples” to “apples” n n For example, the electron reconstruction efficiency can be measured by taking Gsf. Electrons as the probe sample, and setting the passing probe sub-set to be a member of general. Tracks object collection. With this general procedure in mind, the analyzer stores one tag-probe map for each sub-efficiency measurement. The user supplies the set of passing probes and the analyzer stores a pass or fail for each probe in the sample. CSA 08 Electron May 11, 2008 Kalanand Mishra, Fermilab 3
Modular Tag-and-probe software There are 4 modules Tag. And. Probe. Producer - Produces a one-to-many (tag-to-probes) association map. - The tag and probe collections are Candidate objects input by the user. - Input tags for these collections are selected at run time in the configuration file. n Match. Probe. Maker - Produces lists of electron objects that are matched to a different set of electron objects. E. g. , “Gsf. Electrons” matched to “MC truth”. n Tag. Probe. Analyzer - An analyzer that takes the tag-probe association map and makes an eventbased TTree object, which is then written to an output file. - Also stores other interesting objects in the event such as true MC electrons, tracks, HLT / L 1 triggers and kinematic variables associated with them. n Histogrammer - Takes the TTree created by the Tag. Probe. Analyzer and produces plots and makes binned efficiency calculations. - The number of passed / failed signal and background events can be obtained in two different ways: (1. ) by fitting the Z mass peak using likelihood fit with analytical shapes for signal & background. (2. ) by performing a background subtraction in signal region using observed background from the sidebands. n CSA 08 Electron May 11, 2008 Kalanand Mishra, Fermilab 4
Efficiency measurement procedure Likelihood fit for “passing” and “failing” probes Pass Fail y tio All probe n l n o ra t s u Fo CSA 08 Electron May 11, 2008 ll i r Kalanand Mishra, Fermilab 5
An example of efficiency measurement Tracking efficiency: The efficiency of a probe Gsf. Electron to be a member of the general. Tracks collection as function of p. T and . p. T (Ge. V/c) Currently using the following binning scheme (but this is configurable): p. T: [ 0 -20, 20 -40, 40 -60, 60 -80, 80 -100 ] : [ (-2. 5, -1. 5), (-1. 5, 0. 5), (-0. 5, 0. 5), (0. 5, 1. 5), (1. 5, 2. 5) ] CSA 08 Electron May 11, 2008 Kalanand Mishra, Fermilab 6
Example of efficiency measurement continued … You also get the summary of the binned efficiency measurements (in bins of p. T and ) at the end. The result can also be saved in a text file. n Pass condition (Pt): (Probe. Pass == 1 && Pt > 20 && Pt < 40 ) Fail condition (Pt): (Probe. Pass == 0 && Pt > 20 && Pt < 40 ) Num pass 302. 149 Num fail 16. 068 Eff 0. 94932 Eff error 0. 0074461 n ……. n n Pass condition ( Eta ): (Probe. Pass == 1 && Eta > -0. 5 && Eta < 0. 5 ) Fail condition ( Eta ): (Probe. Pass == 0 && Eta > -0. 5 && Eta < 0. 5 ) Num pass 1294. 69 Num fail 41. 3965 Eff 0. 968923 Eff error 0. 003911281 ……. CSA 08 Electron May 11, 2008 Kalanand Mishra, Fermilab 7
How to run the code in CMSSW_2_0_X (upto X = 6) scramv 1 p CMSSW_2_0_6 n cd CMSSW_2_0_6/src n eval `scramv 1 runtime -[c]sh` n cvs co -r V 01 -01 -03 Muon. Analysis/Tag. And. Probe n Download the package “Analysis. Data. Formats/Tag. And. Probe” (needed for tag-probe map dictionary, is not in the CVS yet) : https: //twiki. cern. ch/twiki/bin/viewfile/CMS/Muon. Tag. And. Probe? rev=1; file name=Cand. Tag. And. Probe. Extra. tar. gz n n tar zxvf Cand. Tag. And. Probe. Extra. tar. gz scramv 1 b cd Muon. Analysis/Tag. And. Probe/test 1. First make TTree containing tag-probe collections: cms. Run Analyze. cfg 2. Now calculate efficiency and make efficiency plots: cms. Run electron-efficiency-test. cfg You may need to change the above two configuration files if you want to make your own custom tag-probe pair and want to configure accordingly. CSA 08 Electron May 11, 2008 Kalanand Mishra, Fermilab 8
Things to do …. n n Support efficiency tables for all the officially supported electron Id selectors and startup / expected L 1/HLT conditions. Extend the tag-probe method to [ , J/ , …] e+e-. Done ! Now resonance mass is a configurable parameter. n n n Make sure that the error calculations are done correctly. If the tool becomes useful to analysts, think of ways to estimate systematic uncertainty. Other suggestions …. CSA 08 Electron May 11, 2008 Kalanand Mishra, Fermilab 9
Status & Summary ü A generic tag-and-probe tool for electron and muon efficiency developed ü Method to count the number of signal events can be configured by user: likelihood fit or background subtraction ü Configurable for any combination of tag-probe pair and p. T- binning ü Plan to support efficiency for officially supported electron Id selectors ü Currently available in release CMSSW_2_0_X and 1_8_X ü Your feedbacks are most welcome Thank You ! CSA 08 Electron May 11, 2008 Kalanand Mishra, Fermilab 10
- Slides: 10