Slide 1Slide 2Slide 3Slide 4Slide 5Slide 6Slide 7Slide 8Slide 9Slide 10Slide 11Slide 12Slide 13Slide 14Slide 15Slide 16Slide 17Slide 18Slide 19Slide 20Slide 21Slide 22Slide 23Slide 24Slide 258.882 LHC PhysicsExperimental Methods and MeasurementsAnalysis Tips[Lecture 9, March 4, 2009]‘09The Physics Colloquium SeriesThursday, March 5 at 4:15 pm in room 10-250Marc KamionkowskiCalifornia Institute of Technology"Is the Universe Homogeneous and Isotropic? " SpringFor a full listing of this semester’s colloquia, please visit our website at web.mit.edu/physicsColloquium SeriesPhysicsOrganizational IssuesDuedaysforthedocumentedanalyses●duedatefor'ChargeMultiplicity'isMarch12●worktogetherandhandinoneco-producedassignment●LatexstubhasbeenputontoourcoursepageC.Paus,LHCPhysics:AnalysisTips 4Lecture OutlineCodingexamples●howtoimplementayieldstudy,asanexampleofdesign●applicationtoNTrksanalysis?MatchMonteCarlotruthwithreconstruction●simplekinematicmatch●fullblownhitmatchingSomeDataMonteCarlocomparisons....WhatwouldIexpectashand-in?C.Paus,LHCPhysics:AnalysisTips 5Implementation of a Yield Study Yieldstudy*●determinenumberofcandidatesperlumiandperrun●canbeusedinmanydifferentways●determinewhethertheyieldofsomebackgroundperrun/luminosityisconstant●implementsimplesidebandsubtractionanddeterminewhethersignalyieldisconstantperrun/lumi●identifypotentiallyincompleterunsetc.●likelycandidatetobeusedinvariousanalyses(generic)●perfectopportunitytospendsometimeanddecomposetheproblemDescribingdesignprocessinthefollowingslides!*takenfromBottomMods/toolsareaC.Paus,LHCPhysics:AnalysisTips 6Yield Study DesignStepzero–worthtospenttime?●consider:iscodelikelytobeusedinvariousplaces?●willbeusedallovertheplace●perfectexampleforreducingcomplexityandavoidrepetition●isitgenericandwhatdoesitdependon?●yieldstudycanbeusedforanytypeofselection●studyisabunchofyields●yieldcontains:runNumber,nEvts,intLumi●lookslikeitiscompletelyindependent:noroot,noBStntupleetc.●oneofthoserarecasesC.Paus,LHCPhysics:AnalysisTips 7Yield Study DesignStepone–whatshouldtheusersee?●determinetypicalapplicationinmoredetail●perrun:entrywithrunnumberandluminosity→BeginRun()●perevent:incrementcountdependingonselection→Event(intiE)●atthejobend:showsummary,perrun,maybeshowgraphically→EndJob()●determinetheAPI(ApplicationProgrammerInterface)●YieldStudy(constchar*name)●voidSetRunNumber(intRunNumber,floatLumi)●voidIncrement()●voidPrint(intPrintLevel=0)●intFillArrays(float*Idx,float*Yields,float*Errors)●[workwithseveralyieldstudies:voidAdd(YieldStudy&rhs)]●minimizewhatusersees,withoutaffectingeffectivenessoftoolsC.Paus,LHCPhysics:AnalysisTips 8Yield Study DesignSteptwo–furtherdecompositionuseful?●analyzepiecesandjudge●basicallywehavealist●listentriesaresomewhatmorecomplex:yields●implementyieldclass●followstepsforeachcomponentuntileverythinglayedoutYieldclass●floatValue()returnnEvt/lumi●intRunNumber();voidSetRunNumber(intRunN)●intNEvents();voidSetNEvents(intNEvt)●floatLumi();voidSetLumi(floatLumi)●voidPrint()●voidIncNEvents()C.Paus,LHCPhysics:AnalysisTips 9Yield Study DesignDetailsprogrammercanhidefromtheuser●runnumbersarenotcontiguous●implementationneedstokeepseparateaccountofeachrunnumberthatappearedandonlycreatenewentryifnumberhasnotyetappeared●luminosityisstoredeachtimefortheentirerun(noaddingneeded)●uncertaintiescanbeintrinsicallyhandledonthebasisofGaussianorPoissonianstatistics(weuseGaussian)●implementationofbunchofyieldscouldbe:anarray,list,vectorormap,whateverismostappropriate●errorhandlingforinvalidluminositiesofeventnumbers●printlevelcanbetunedtoonesneedsPossibleextensions:combineYieldStudies+/--C.Paus,LHCPhysics:AnalysisTips 10Yield Study: An ImplementationC.Paus,LHCPhysics:AnalysisTips 11Use in NTrks Analysis?Implementedselectionofevents●ifnotrackinside|η|<0.5●expectation:someeventsshouldbeselected●yield(nEvent/Lumi)shouldbeconstant●isthisstudyuseful?yes!●therecouldbe'bad'runs(trackeroff,noeventhastracks)●couldbearunwasincompletelyprocessed(yieldtoosmall)●anythingwecannotthinkofrightnow●butcareful:randomtestsofyoursamplemightbiasyourviewonwhatisgoingon●tofixaproblem,like'wasthetrackeroffinthisrun?':cannotuseyieldasargument●wehavetofindtherealreasonfortheproblemandfixitindependentlyofouranalysissampleC.Paus,LHCPhysics:AnalysisTips 12Yield Study Result in NTrksConclusion:somethingiswronghere!●expectedyieldtobeconstantbutseewildvariations?C.Paus,LHCPhysics:AnalysisTips 13NTrks