Objects may be reordered rounded or resized they may have names or attributes removed or they may even be coerced to a new class if necessary in order to achieve equality The results of comparisons report not just whether the objects are the same bu ID: 8743
Download Pdf The PPT/PDF document "Comparing NonIdentical Objects Introduci..." is the property of its rightful owner. Permission is granted to download and print the materials on this web site for personal, non-commercial use only, and to display it on your personal computer provided you do not modify the materials and that you retain all copyright notices contained in the materials. By downloading content from our website, you accept the terms of this agreement.
1.WriteRcodetocreatethethreevectorsandthefactorshownbelow,withnamesid,age,edu,andclass.Youshouldendupwithob-jectsthatlooklikethis:�id[1]123456�age[1]303228392025�edu[1]000000�class[1]poorpoorpoormiddle[5]middlemiddleLevels:middlepoor2.CombinetheobjectsfromQuestion1togethertomakeadataframecalledIndianMothers.Youshouldendupwithanobjectthatlookslikethis:�IndianMothersidageeduclass11300poor22320poor33280poor44390middle55200middle66250middle Figure1:TwosimpleexamplesoftheexercisesthatSTATS220studentsareaskedtoperform.ThisreturnsTRUEifthetwoobjectsareexactlythesame,otherwiseitreturnsFALSE.Theproblemwiththisfunctionisthatitisverystrictindeedandwillfailforobjectsthatare,forallpracticalpurposes,thesame.Theclassicexampleisthecomparisonoftworeal( oating-point)values,asdemonstratedinthefollowingcode,wheredi erencescanarisesimplyduetothelimitationsofhownumbersarerepresentedincomputermemory(seeRFAQ7.31,Hornik,2008).�identical(0.3-0.2,0.1)[1]FALSEUsingthefunctiontotestforequalitywouldclearlybeunreasonablyharshwhenmarkinganystudentanswerthatinvolvescalculatinganumericresult.Theidentical()function,byitself,isnotsucientforcomparingstudentanswerswithmodelanswers.ShadesofgreyTherecommendedsolutiontotheproblemmentionedaboveofcomparingtwo oating-pointvaluesistousetheall.equal()function.Thisfunctionallowsfor\insignicant"di erencesbetweennumericvalues,asshownbelow.�all.equal(0.3-0.2,0.1)[1]TRUE2 Thismakesall.equal()amuchmoreappropriatefunctionforcomparingstu-dentanswerswithmodelanswers.Whatislesswell-knownabouttheall.equal()functionisthatitalsoworksforcomparingothersortsofRobjects,besidesnumericvectors,andthatitdoesmorethanjustreportequalitybetweentwoobjects.Iftheobjectsbeingcomparedhavedi erences,thenall.equal()doesnotsimplyreturnFALSE.Instead,itreturnsacharactervectorcontainingmessagesthatdescribethedi erencesbetweentheobjects.Thefollowingcodegivesasimpleexample,whereall.equal()reportsthatthetwocharactervectorshavedi erentlengths,andthat,ofthetwopairsofstringsthatcanbecompared,onepairofstringsdoesnotmatch.�all.equal(c("a","b","c"),c("a","B"))[1]"Lengths(3,2)differ(stringcompareonfirst2)"[2]"1stringmismatch"Thisfeatureisactuallyveryusefulformarkingstudentwork.Informationaboutwhetherastudent'sansweriscorrectisusefulfordeterminingarawmark,butitisalsousefultohaveinformationaboutwhatthestudentdidwrong.Thisinformationcanbeusedasthebasisforassigningpartialmarksforananswerthatisclosetothecorrectanswer,andforprovidingfeedbacktothestudentaboutwheremarkswerelost.Theall.equal()functionhassomeusefulfeaturesthatmakeitahelpfultoolforcomparingstudentanswerswithmodelanswers.However,thereisanapproachthatcanperformbetterthanthis.Theall.equal()functionlooksforequalitybetweentwoobjectsand,ifthatfails,providesinformationaboutthesortofdi erencesthatexist.Analternativeapproach,whentwoobjectsarenotequal,istotrytotransformtheobjectstomakethemequal,andreportonwhichtransformationswerenecessaryinordertoachieveequality.Asanexampleofthedi erencebetweentheseapproaches,considerthetwoobjectsbelow:acharactervectorandafactor.�obj1-c("a","a","b","c")ကobj1[1]"a""a""b""c"ကobj2-factor(obj1)ကobj2[1]aabcLevels:abcTheall.equal()functionreportsthattheseobjectsaredi erentbecausetheydi erintermsoftheirfundamentalmode|onehasattributesandtheotherdoesnot|andbecauseeachobjectisofadi erentclass.ကall.equal(obj1,obj2)[1]"Modes:character,numeric"[2]"Attributes:targetisNULL,currentislist&#x-531;"[3]"targetischaracter,currentisfactor"3 Thealternativeapproachwouldbetoallowvarioustransformationsoftheobjectstoseeiftheycanbetransformedtobethesame.Thefollowingcodeshowsthisapproach,whichreportsthattheobjectsareequal,ifthesecondoneiscoercedfromafactortoacharactervector.Thisismoreinformationthanwasprovidedbyall.equal()and,intheparticularcaseofcomparingstudentanswerstomodelanswers,ittellsusalotabouthowclosethestudentgottotherightanswer.�library(compare)�compare(obj1,obj2,allowAll=TRUE)TRUEcoercedfromfactorကtocharacterကAnotherlimitationofall.equal()isthatitdoesnotreportonsomeotherpossibledi erencesbetweenobjects.Forexample,itispossibleforastudenttohavethecorrectvaluesforanRobject,buthavethevaluesinthewrongorder.Anothercommonmistakeistogetthecasewronginasetofstringvalues(e.g.,inacharactervectororinthenamesattributeofanobject).Insummary,whileall.equal()providessomedesirablefeaturesforcom-paringstudentanswerstomodelanswers,wecandobetterbyallowingforawiderrangeofdi erencesbetweenobjectsandbytakingadi erentapproachthatattemptstotransformthestudentanswertobethesameasthemodelanswer,ifatallpossible,whilereportingwhichtransformationswerenecessary.Theremainderofthisarticledescribesthecomparepackage,whichprovidesfunctionsforproducingthesesortsofcomparisons.Thecompare()functionThemainfunctioninthecomparepackageisthecompare()function.Thisfunctioncheckswhethertwoobjectsarethesameand,iftheyarenot,carriesoutvarioustransformationsontheobjectsandchecksthemagaintoseeiftheyarethesameaftertheyhavebeentransformed.Bydefault,compare()onlysucceedsifthetwoobjectsareidentical(usingtheidentical()function)orthetwoobjectsarenumericandtheyareequal(accordingtoall.equal()).Iftheobjectsarenotthesame,notransformationsoftheobjectsareconsidered.Inotherwords,bydefault,compare()issimplyaconveniencewrapperforidentical()andall.equal().Asasimpleexam-ple,thefollowingcomparisontakesaccountofthefactthatthevaluesbeingcomparedarenumericandusesall.equal()ratherthanidentical().ကcompare(0.3-0.2,0.1)TRUETransformationsThemoreinterestingusesofcompare()involvespecifyingoneormoreoftheargumentsthatallowtransformationsoftheobjectsthatarebeingcompared.Forexample,thecoerceargumentspeciesthatthesecondargumentmaybecoercedtotheclassoftherstargument.Thisallowsformore exiblecomparisonssuchasbetweenafactorandacharactervector.4 �compare(obj1,obj2,coerce=TRUE)TRUEcoercedfromfactorကtocharacterကItisimportanttonotethatthereisadeniteordertotheobjects;themodelobjectisgivenrstandthecomparisonobjectisgivensecond.Transformationsattempttomakethecomparisonobjectlikethemodelobject,thoughinanum-berofcases(e.g.,whenignoringthecaseofstrings)themodelobjectmayalsobetransformed.Intheexampleabove,thecomparisonobjecthasbeencoercedtobethesameclassasthemodelobject.Thefollowingcodedemonstratesthee ectofreversingtheorderoftheobjectsinthecomparison.Nowthecharactervectorisbeingcoercedtoafactor.ကcompare(obj2,obj1,coerce=TRUE)TRUEcoercedfromcharacterကtofactorကOfcourse,transforminganobjectisnotguaranteedtoproduceidenticalobjectsiftheoriginalobjectsaregenuinelydi erent.ကcompare(obj1,obj2[1:3],coerce=TRUE)FALSEcoercedfromfactorကtocharacterကNotice,however,thateventhoughthecomparisonfailed,theresultstillreportsthetransformationthatwasattempted.Thisresultindicatesthatthecompar-isonobjectwasconvertedfromafactor(toacharactervector),butitstilldidnotendupbeingthesameasthemodelobject.Anumberofothertransformationsareavailableinadditiontocoercion.Forexample,di erencesinlength,likeinthelastcase,canalsobeignored.ကcompare(obj1,obj2[1:3],+shorten=TRUE,coerce=TRUE)TRUEcoercedfromfactorကtocharacterကshortenedmodelItisalsopossibletoallowvaluestobesorted,orrounded,ortoconvertallcharactervaluestouppercase(i.e.,ignorethecaseofstrings).Table1providesacompletelistofthetransformationsthatarecurrentlyallowed(inversion0.2ofcompare)andtheargumentsthatareusedtoenablethem.Afurtherargumenttothecompare()function,allowAll,controlsthede-faultsettingformostofthesetransformations,sospecifyingallowAll=TRUEisaquickwayofenablingallpossibletransformations.SpecictransformationscanstillbeexcludedbyexplicitlysettingtheappropriateargumenttoFALSE.TheequalargumentisabitofaspecialcasebecauseitisTRUEbydefault,whereasalmostallothersareFALSE.Theequalargumentisalsoespecially5 in uentialbecauseobjectsarecomparedaftereverytransformationandthisargumentcontrolswhatsortofcomparisontakesplace.Objectsarealwayscomparedusingidentical()rst,whichwillonlysucceediftheobjectshaveexactlythesamerepresentationinmemory.Ifthetestusingidentical()failsandequal=TRUE,thenamorelenientcomparisonisalsoperformed.Bydefault,thisjustmeansthatnumericvaluesarecomparedusingall.equal(),butvariousotherargumentscanextendthistoallowthingslikedi erencesincaseforcharactervalues(seetheasteriskedargumentsinTable1).TheroundargumentisalsospecialbecauseitalwaysdefaultstoFALSE,evenifallowAll=TRUE.Thismeansthattheroundargumentmustbespeciedexplicitlyinordertoenablerounding.ThedefaultissetupthiswaybecausethevalueoftheroundargumentiseitherFALSEoranintegervaluespecifyingthenumberofdecimalplacestoroundto.Forthisargument,thevalueTRUEcorrespondstoroundingtozerodecimalplaces.Finally,thereisanadditionalargumentcolsOnlyforcomparingdataframes.Thisargumentcontrolswhethertransformationsareonlyappliedtocolumns(andnottorows).Forexample,bydefault,adataframewillonlyallowcolumnstobedropped,butnotrows,ifshorten=TRUE.Note,however,thatignoreOrdermeansignoretheorderofrowsfordataframesandignoreColOrdermustbeusedtoignoretheorderofcolumnsincomparisonsinvolvingdataframes.ThecompareName()functionThecompareName()functiono ersaslightvariationonthecompare()function.Forthisfunction,onlythenameofthecomparisonobjectisspecied,ratherthananexplicitobject.Theadvantageofthisisthatitallowsforvariationsincaseinthenamesofobjects.Forexample,astudentmightcreateavari-ablecalledindianMothersratherthanthedesiredIndianMothers.Thiscase-insensitivityisenabledviatheignore.caseargument.Anotheradvantageofthisfunctionisthatitispossibletospecify,viathecompEnvargument,aparticularenvironmenttosearchwithinforthecomparisonobject(ratherthanjustthecurrentworkspace).Thisbecomesusefulwhencheckingtheanswersfromseveralstudentsbecauseeachstudent'sanswersmaybegeneratedwithinaseparateenvironmentinordertoavoidanyinteractionsbetweencodefromdi erentstudents.Thefollowingcodeshowsasimpledemonstrationofthisfunction,whereacomparisonobjectiscreatedwithinatemporaryenvironmentandthenameofthecomparisonobjectisuppercasewhenitshouldbelowercase.�tempEnv-new.env()ကwith(tempEnv,X-1:10)ကcompareName(1:10,"x",compEnv=tempEnv)TRUErenamedobjectNoticethat,aswiththetransformationsincompare(),thecompareName()func-tionrecordswhetheritneededtoignorethecaseofthenameofthecomparisonobject.7 ApathologicalexampleThissectionshowsamanufacturedexamplethatdemonstratessomeofthe exibilityofthecompare()function.Wewillcomparetwodataframesthathaveanumberofsimpledi erences.Themodelobjectisadataframewiththreecolumns:anumericvector,acharactervector,andafactor.�model-+data.frame(x=1:26,+y=letters,+z=factor(letters),+row.names=letters,+stringsAsFactors=FALSE)Thecomparisonobjectcontainsessentiallythesameinformation,exceptthatthereisanextracolumn,thecolumnnamesareuppercaseratherthanlowercase,thecolumnsareinadi erentorder,theyvariableisafactorratherthanacharactervector,andthezvariableisacharactervariableratherthanafactor.Theyvariableandtherownamesarealsouppercaseratherthanlowercase.ကcomparison-+data.frame(W=26:1,+Z=letters,+Y=factor(LETTERS),+X=1:26,+row.names=LETTERS,+stringsAsFactors=FALSE)Thecompare()functioncandetectthatthesetwoobjectsareessentiallythesameaslongaswereorderthecolumns(ignoringthecaseofthecolumnnames),coercetheyandzvariables,droptheextravariable,ignorethecaseoftheyvariable,andignorethecaseoftherownames.ကcompare(model,comparison,allowAll=TRUE)TRUErenamedreorderedcolumns[Y]coercedfromfactorကtocharacterက[Z]coercedfromcharacterကtofactorကshortenedcomparison[Y]ignoredcaserenamedrowsNoticethatwehaveusedallowAll=TRUEtoallowcompare()toattemptallpossibletransformationsatitsdisposal.ComparinglesofRcodeReturningnowtotheoriginalmotivationforthecomparepackage,thecompare()functionprovidesanexcellentbasisfordeterminingnotonlywhetherastudent's8 answersarecorrect,butalsohowmuchincorrectanswersdi erfromthemodelanswer.Asdescribedearlier,submissionsbystudentsintheSTATS220coursecon-sistoflesofRcode.Markingthesesubmissionsconsistsofusingsource()torunthecode,thencomparingtheresultingobjectswithmodelanswerobjects.Withapproximately100studentsintheSTATS220course,withweeklylabs,andwithmultiplequestionsperlab,eachofwhichmaycontainmorethanoneRobject,thereisareasonablemarkingburden.Consequently,thereisastrongincentivetoautomateasmuchofthemarkingprocessaspossible.ThecompareFile()functionThecompareFile()functioncanbeusedtorunRcodefromaspecicleandcomparetheresultswithasetofmodelanswers.Thisfunctionrequiresthreepiecesofinformation:thenameofalecontainingthe\comparisoncode",whichisrunwithinalocalenvironment,usingsource(),togeneratethecomparisonvalues;avectorof\modelnames",whicharethenamesoftheobjectsthatwillbelookedforinthelocalenvironmentafterthecomparisoncodehasbeenrun;andthemodelanswers,eitherasthenameofabinaryletoload(),orasthenameofaleofRcodetosource(),orasalistobjectcontainingtheready-mademodelanswerobjects.Anyargumenttocompare()mayalsobeincludedinthecall.Oncethecomparisoncodehasbeenrun,compareName()iscalledforeachofthemodelnamesandtheresultisalistof"comparison"objects.Asasimpledemonstration,considerthebasicquestionsshowninFigure1.Themodelnamesinthiscasearethefollowing:�modelNames-c("id","age",+"edu","class",+"IndianMothers")Onestudent'ssubmissionforthisexerciseisinalecalledstudent1.R,withinadirectorycalledExamples.Themodelanswerisinalecalledmodel.Rinthesamedirectory.Wecanevaluatethisstudent'ssubmissionandcompareittothemodelanswerwiththefollowingcode:ကcompareFile(file.path("Examples",+"student1.R"),+modelNames,+file.path("Examples",+"model.R"))$idTRUE$ageTRUE$eduTRUE9 $classFALSE$IndianMothersFALSEobjectnotfoundThisprovidesastrictcheckandshowsthatthestudentgottherstthreeproblemscorrect,butthelasttwowrong.Infact,thestudent'scodecompletelyfailedtogenerateanobjectwiththenameIndianMothers.Wecanprovideextraargumentstoallowtransformationsofthestudent'sanswers,asinthefollowingcode:�compareFile(file.path("Examples",+"student1.R"),+modelNames,+file.path("Examples",+"model.R"),+allowAll=TRUE)$idTRUE$ageTRUE$eduTRUE$classTRUEreorderedlevels$IndianMothersFALSEobjectnotfoundThisshowsthat,althoughthestudent'sanswerfortheclassobjectwasnotperfect,itwasprettyclose;itjusthadthelevelsofthefactorinthewrongorder.ThecompareFiles()functionThecompareFiles()functionbuildsoncompareFile()byallowingavectorofcomparisonlenames.Thisallowsawholesetofstudentsubmissionstobetestedatonce.Theresultofthisfunctionisalistoflistsof"comparison"objectsandaspecialprintmethodprovidesasimpliedviewofthisresult.Continuingtheexamplefromabove,theExamplesdirectorycontainssub-missionsfromafurtherfourstudents.WecancompareallofthesesubmissionswiththemodelanswersandproduceasummaryoftheresultswithasinglecalltocompareFiles().TheappropriatecodeandoutputareshowninFigure2.10 �files-list.files("Examples",+pattern="^student[0-9]+[.]R$",+full.names=TRUE)ကresults-compareFiles(files,+modelNames,+file.path("Examples","model.R"),+allowAll=TRUE,+resultNames=gsub("Examples.|[.]R","",files))ကresultsidageeduclassIndianMothersstudent1TRUETRUETRUETRUEreorderedlevelsFALSEobjectnotfoundstudent2TRUETRUETRUETRUETRUEstudent3TRUETRUETRUETRUEcoercedfromcharacterကtofactorကFALSEobjectnotfoundstudent4TRUETRUETRUETRUEcoercedfromcharacterကtofactorကTRUErenamedobjectstudent5TRUETRUETRUEFALSEobjectnotfoundFALSEobjectnotfoundFigure2:UsingthecompareFiles()functiontorunRcodefromseverallesandcomparetheresultstomodelobjects.Theresultofthissortofcompar-isoncaneasilygetquitewide,soitisoftenusefultoprinttheresultwithoptions(width)settosomelargevalueandusingasmallfont,ashasbeendonehere.Theresultsshowthatmoststudentsgottherstthreeproblemscorrect.Theyhadmoretroublegettingthefourthproblemright,withonegettingthefactorlevelsinthewrongorderandtwoothersproducingacharactervectorratherthanafactor.Onlyonestudent,student2,gotthenalproblemexactlyrightandonlyoneother,student4,gotessentiallytherightanswer,thoughthisstudentspeltthenameoftheobjectwrong.AssigningmarksandgivingfeedbackTheresultreturnedbycompareFiles()isalistoflistsofcomparisonresults,whereeachresultisitselfalistofinformationincludingwhethertwoobjectsarethesameandarecordofhowtheobjectsweretransformedduringthecomparison.ThisrepresentsawealthofinformationwithwhichtoassesstheperformanceofstudentsonasetofRexercises,butitcanbealittleunwieldlytodealwith.Thecomparepackageprovidesfurtherfunctionsthatmakeiteasiertodealwiththisinformationforthepurposeofdetermininganalmarkandforthepurposeofprovidingcommentsforeachstudentsubmission.Inordertodetermineanalmark,weusethequestionMarks()functiontospecifywhichobjectnamesareinvolvedinaparticularquestion,toprovideamaximummarkforthequestion,andtospecifyasetofrulesthatdeterminehowmanymarksshouldbedeductedforvariousdeviationsfromthecorrectanswers.Therule()functionisusedtodeneamarkingrule.Ittakesanobjectname,anumberofmarkstodeductifthecomparisonforthatobjectisFALSE,11 plusanynumberoftransformationrules.ThelatteraregeneratedusingthetransformRule()function,whichassociatesaregularexpressionwithanum-berofmarkstodeduct.Iftheregularexpressionismatchedintherecordoftransformationsforacomparison,thentheappropriatenumberofmarksarededucted.Asimpleexample,basedonthesecondquestioninFigure1,isshownbelow.ThisspeciesthatthequestiononlyinvolvesanobjectnamedIndianMothers,thatthereisamaximummarkof1forthisquestion,andthat1markisdeductedifthecomparisonisFALSE.�q2-+questionMarks("IndianMothers",+maxMark=1,+rule("IndianMothers",1))TherstquestionfromFigure1providesamorecomplexexample.Inthiscase,therearefourdi erentobjectsinvolvedandthemaximummarkis2.TherulesbelowspecifythatanyFALSEcomparisondropsamarkandthat,forthecomparisoninvolvingtheobjectnamed"class",amarkshouldalsobedeductedifcoercionwasnecessarytogetaTRUEresult.ကq1-+questionMarks(+c("id","age","edu","class"),+maxMark=2,+rule("id",1),+rule("age",1),+rule("edu",1),+rule("class",1,+transformRule("coerced",1)))Havingsetupthismarkingscheme,marksaregeneratedusingthemarkQuestions()function,asshownbythefollowingcode.ကmarkQuestions(results,q1,q2)id-age-edu-classIndianMothersstudent120student221student310student411student510Fortherstquestion,thethirdandfourthstudentsloseamarkbecauseofthecoercion,andthefthstudentlosesamarkbecausehehasnotgeneratedtherequiredobject.Asimilarsuiteoffunctionsareprovidedtoassociatecomments,ratherthanmarkdeductions,withparticulartransformations.Thefollowingcodeprovidesasimpledemonstration.ကq1comments-+questionComments(12 +c("id","age","edu","class"),+comments(+"class",+transformComment(+"coerced",+"'class'isafactor!")))�commentQuestions(results,q1comments)id-age-edu-classstudent1""student2""student3"'class'isafactor!"student4"'class'isafactor!"student5""Inthiscase,wehavejustgeneratedfeedbackforthestudentswhogeneratedacharactervectorinsteadofthedesiredfactorinQuestion1oftheexercise.Summary,discussion,andfuturedirectionsThecomparepackageisbasedaroundthecompare()function,whichcomparestwoobjectsforequalityand,iftheyarenotequal,attemptstotransformtheobjectstomakethemequal.Itreportswhetherthecomparisonsucceededoverallandprovidesarecordofthetransformationsthatwereattemptedduringthecomparison.Furtherfunctionsareprovidedontopofthecompare()functiontofacilitatemarkingexerciseswherestudentsinaclasssubmitRcodeinaletocreateasetofRobjects.Thisarticlehasgivensomebasicdemonstrationsoftheuseofthecompare()packageforcomparingobjectsandmarkingstudentsubmissions.Thepackagecouldalsobeusefulforthestudentsthemselves,bothtocheckwhethertheyhavethecorrectanswerandtoprovidefeedbackabouthowtheiranswerdi ersfromthemodelanswer.Moregenerally,thecompare()functionmayhaveapplicationwherevertheidentical()andall.equal()functionsarecurrentlyinuse.Forexample,itmaybeusefulwhendebuggingcodeandforperformingregressiontestsaspartofaqualitycontrolprocess.Obviousextensionsofthecomparepackageincludeaddingnewtransfor-mationsandprovidingcomparisonmethodsforotherclassesofobjects.Moredetailsabouthowthepackageworksandhowtheseextensionsmightbedevel-opedarediscussedinthevignette,\FundamentalsoftheComparePackage",whichisinstalledaspartofthecomparepackage.AcknowledgementsManythankstotheeditorsandanonymousreviewersfortheirusefulcommentsandsuggestions,onboththisarticleandthecomparepackageitself.13 ReferencesK.~Hornik.TheRFAQ,2008.URLhttp://CRAN.R-project.org/doc/FAQ/R-FAQ.html.ISBN3-900051-08-9.RDevelopmentCoreTeam.R:ALanguageandEnvironmentforStatisticalComputing.RFoundationforStatisticalComputing,Vienna,Austria,2008.URLhttp://www.R-project.org.ISBN3-900051-07-0.14