Professional Documents
Culture Documents
AssignmentiNo:i02
Name:iSeeratiFatima
iiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiRolliNo:iBY674855
iiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiCourse:iEducationaliStatisticsi(Professional)
iiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiCourseiCode:i8614
iiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiSubmittedito:ikausar Batool
Questionino:i01:-iDefineihypothesisitestingiandilogicibehindihypothesisitesting.
Answer:i
HypothesisiTestingiItiisiusuallyiimpossibleiforiairesearcheritoiobserveieachiindividualiiniaip
opulation.iTherefore,iheiselectsisomeiindividualifromitheipopulationiasisampleiandicollects
idataifromitheisample.iHeitheniusesitheisampleidataitoiansweriquestionsiaboutitheipopula
tion.iForithisipurpose,iheiusesisomeistatisticalitechniques.iHypothesisitestingiisiaistatisticali
methodithatiusesisampleidataitoievaluateiaihypothesisiaboutiaipopulationiparameteri(Grav
etteri&iWallnau,i2002).Aihypothesisitestiisiusuallyiusediinicontextiofiairesearchistudy.iDep
endingionitheitypeiofiresearchianditheitypeiofidata,itheidetailsiofitheihypothesisitestiwillic
hangeifromionisituationitoianother.iHypothesisitestingiisiaiformalizediprocedureithatifollo
wsiaistandardiseriesiofioperations.iInithisiwayiairesearcherihasiaistandardizedimethodiforie
valuatingitheiresultsiofihisiresearchistudy.iOtheriresearchersiwillirecognizeiandiunderstandi
exactlyihowitheidataiwereievaluatediandihowiconclusionsiwereidrawn.
LogiciofiHypothesisiTestingi
AccordingitoiGravetteri&iWallnaui(2002)itheilogiciunderlyingihypothesisitestingiisiasifollow
s:i
2
i) First,iairesearcheristatesiaihypothesisiaboutiaipopulation.iUsually,itheihypothesisicon
cernsitheivalueiofitheipopulationimean.iForiexample,iweimightihypothesizeithatithei
meaniIQiforitheiregisteredivotersiPakistaniisiMi=i100.i
ii) iBeforeiairesearcheriactuallyiselectsiaisample,iheiusesitheihypothesisitoipredictitheich
aracteristicsithatitheisampleishouldihave.iForiexample,iifiheihypothesizesithatitheipo
pulationimeaniIQi=i100,itheniheiwouldipredictithatitheisampleishouldihaveiaimeaniar
oundi100.iItishouldibeikeptiinimindithatitheisampleishouldibeisimilaritoitheipopulatio
nibutithereiisialwaysiaichanceicertainiamountiofierror.i
iii) iNext,itheiresearcheriobtainsiairandomisampleifromitheipopulation.iForiexample,ihei
mightiselectiairandomisampleiofini=i200iregisteredivotersitoicomputeitheimeaniIQifo
ritheisample.i
iv) Finally,iheicomparesitheiobtainedisampleidataiwithitheipredictionithatiwasimadeifro
mitheihypothesis.iIfitheisampleimeaniisiconsistentiwithitheiprediction,iheiwilliconclud
eithatitheihypothesisiisireasonable.iButiifithereiisibigidifferenceibetweenitheidataiand
itheiprediction,iheiwillidecideithatitheihypothesisiisiwrong.
Four-StepiProcessiforiHypothesisiTestingi
Theiprocessiofihypothesisitestingigoesithroughifollowingifouristeps.i
i) StatingitheiHypothesisiTheiprocessiofihypothesisitestingibeginsibyistatingiaihypoth
esisiaboutitheiunknownipopulation.iUsually,iairesearcheristatesitwoiopposingihypot
heses.iAndibothihypothesesiareistatediinitermsiofipopulationiparameters.iTheifirsti
andimostiimportantiofitwoihypothesesiisicalledinullihypothesis.iAinullihypothesisist
atesithatitheitreatmentihasinoieffect.iInigeneral,inullihypothesisistatesithatithereiisi
noichange,inoieffect,inoidifferencei–
inothingihappened.iTheinullihypothesisiisidenotedibyitheisymboliHoi(Histandsiforih
ypothesisiandi0idenotesithatithisiisizeroieffect).iTheinullihypothesisi(Ho)istatesithati
initheigeneralipopulationithereiisinoichange,inoidifference,iorinoirelationship.iIniani
experimentalistudy,inullihypothesisi(Ho)ipredictsithatitheiindependentivariablei(tre
atment)iwillihaveinoieffectionitheidependentivariableiforitheipopulation.iTheisecon
3
dihypothesisiisisimplyitheioppositeiofinullihypothesisiandiitiisicalleditheiscientificiori
alternativeihypothesis.iItiisidenotedibyiH1.iThisihypothesisistatesithatitheitreatment
ihasianieffectionitheidependentivariable.iTheialternativeihypothesisi(H1)istatesithati
thereiisiaichange,iaidifference,ioriairelationshipiforitheigeneralipopulation.iInianiex
periment,iH1ipredictsithatitheiindependentivariablei(treatment)iwillihaveianieffecti
onitheidependentivariable.
ii) SettingiCriteriaiforitheiDecisioni
Iniaicommonipractice,iairesearcheriusesitheidataifromitheisampleitoievaluateitheiauthorityiofi
nullihypothesis.iTheidataiwillieitherisupportiorinegateitheinullihypothesis.iToiformalizeitheidec
isioniprocess,iairesearcheriwilliuseinullihypothesisitoipredictiexactlyiwhatikindiofisampleishoul
dibeiobtainediifitheitreatmentihasinoieffect.iIniparticular,iairesearcheriwilliexamineiallitheiposs
ibleisampleimeansithaticouldibeiobtainediifitheinullihypothesisiisitrue.
iii) Collectingidataiandicomputingisampleistatisticsi
Theinextistepiinihypothesisitestingiisitoiobtainitheisampleidata.iThenirawidataiareisummarized
iwithiappropriateistatisticsisuchiasimean,istandardideviationietc.itheniitiisipossibleiforitheirese
archeritoicompareitheisampleimeaniwithitheinullihypothesis.
iv) MakeiaiDecisioni
Initheifinalistepitheiresearcheridecides,iinitheilightiofianalysisiofidata,iwhetheritoiacceptiorirej
ectitheinullihypothesis.iIfianalysisiofidataisupportsitheinullihypothesis,iheiacceptsiitiandiviceive
rsa.i
UncertaintyiandiErroriiniHypothesisi
TestingiHypothesisitestingiisianiinferentialiprocess.iItimeansithatiitiusesilimitediinformationiob
tainedifromitheisampleitoireachigeneraliconclusionsiaboutitheipopulation.iAsiaisampleiisiaisma
llisubsetiofitheipopulation,iitiprovidesionlyilimitedioriincompleteiinformationiaboutitheiwholei
population.iYetihypothesisitestiusesiinformationiobtainedifromitheisample.iInithisisituation,ith
4
ereiisialwaysitheiprobabilityiofireachingiincorrecticonclusion.iGenerallyitwoikindsiofierrorsicani
beimade.
TypeiIiErrorsi
AitypeiIierrorioccursiwheniairesearcherirejectsiainullihypothesisithatiisiactuallyitrue.iItimeansit
hatitheiresearchericoncludesithatitheitreatmentidoesihaveianieffectiwheniinifactitheitreatmen
tihasinoieffect.iTypeiIierroriisinotiaistupidimistakeiinitheisenseithatitheiresearcheriisioverlookin
gisomethingithatishouldibeiperfectlyiobvious.iHeiisilookingiatitheidataiobtainedifromitheisamp
leithatiappearitoishowiaiclearitreatmentieffect.iTheiresearcherithenimakesiaicarefulidecisionib
asedioniavailableiinformation.iHeineveriknowsiwhetheriaihypothesisiisitrueiorifalse.iTheiconse
quencesiofiaitypeiIierroricanibeiveryiseriousibecauseitheiresearcherihasirejecteditheinullihypot
hesisiandibelievedithatitheitreatmentihadiairealieffect.iitiisilikelyithatitheiresearcheriwillireport
ioripublishitheiresearchiresults.iOtheriresearchersimayitryitoibuilditheoriesioridevelopiotheriex
perimentsibasedionifalseiresults.
TypeiIIiErrorsi
AitypeiIIierrorioccursiwheniairesearcherifailsitoirejectitheinullihypothesisithatiisireallyifalse.iIti
meansithatiaitreatmentieffectireallyiexists,ibutitheihypothesisitestihasifaileditoidetectiit.iThisit
ypeiofierrorioccursiwhenitheieffectiofitheitreatmentiisirelativelyismall.iThatiisitheitreatmentid
oesiinfluenceitheisampleibutitheimagnitudeiofitheieffectiisiveryismall.iTheiconsequencesiofiTy
peiIIierroriareinotiveryiserious.i
InicaseiofiTypeiIIierroritheiresearchidataidoinotishowitheiresultsithatitheiresearcherihadihoped
itoiobtain.iTheiresearchericaniacceptithisioutcomeiandiconcludeithatitheitreatmentieitherihasi
noieffectiorihasiaismallieffectithatiisinotiworthipursuing.iOritheiresearchericanirepeatitheiexpe
rimentiwithisomeiimprovementianditryitoidemonstrateithatitheitreatmentidoesiwork.iItiisiimp
ossibleitoidetermineiaisingle,iexactiprobabilityivalueiforiaitypeiIIierror.iSummarizingiweicanisay
ithatiaihypothesisitestialwaysileadsitoioneiofitwoidecisions.ii
Theisampleidataiprovidesisufficientievidenceitoirejectitheinullihypothesisianditheiresea
rchericoncludesithatitheitreatmentihasianieffect.i
5
Theisampleidataidoinotiprovideienoughievidenceitoirejectitheinullihypothesis.iTheirese
archerifailsitoirejectitheinullihypothesisiandiconcludesithatitheitreatmentidoesinotiapp
earitoihaveianieffect.
Inieithericase,ithereiisiaichanceithatitheidataiareimisleadingianditheidecisioniisiwrong.
iiiiiiiiiiiiiiiiiiiiiiii________________________________________________
Questionino:i02:-
iExplainitypesiofiANOVA.iDescribeipossibleisituationsiiniwhichieachitypeishouldibeiused.
Answer:
AnalysisiofiVariancei(ANOVA)iisiaistatisticaliprocedureiuseditoitestitheidegreeitoiwhichitwoiori
moreigroupsivaryioridifferiinianiexperiment.
Theit-testsihaveioneiveryiseriousilimitationi–
itheyiareirestricteditoitestsiofitheisignificanceiofitheidifferenceibetweenionlyitwoigroups.iTher
eiareimanyitimesiwheniweilikeitoiseeiifithereiareisignificantidifferencesiamongithree,ifour,iorie
venimoreigroups.iForiexampleiweimayiwantitoiinvestigateiwhichiofithreeiteachingimethodsiisi
bestiforiteachingininthiclassialgebra.iInisuchicase,iweicannotiuseit-
testibecauseimoreithanitwoigroupsiareiinvolved.iToidealiwithisuchitypeioficasesioneiofitheimo
stiusefulitechniquesiinistatisticsiisianalysisiofivariancei(abbreviatediasiANOVA).iThisitechniquei
wasidevelopedibyiaiBritishiStatisticianiRonaldiA.iFisheri(Dietzi&iKalof,i2009;iBartz,i1981)iAnalys
isiofiVariancei(ANOVA)iisiaihypothesisitestingiprocedureithatiisiuseditoievaluateimeanidifferen
cesibetweenitwoiorimoreitreatmentsi(oripopulation).iLikeialliotheriinferentialiprocedures.iANO
VAiusesisampleidataitoiasiaibasisiforidrawingigeneraliconclusioniaboutipopulations.iSometime,
iitimayiappearithatiANOVAiandit-
testiareitwoidifferentiwaysiofidoingiexactlyisameithing:itestingiforimeanidifferences.iInisomeic
asedithisiisitruei–
ibothitestsiuseisampleidataitoitestihypothesisiaboutipopulationimean.iHowever,iANOVAihasim
6
uchimoreiadvantagesioverit-test.it-
testsiareiusediwheniweihaveicompareionlyitwoigroupsiorivariablesi(oneiindependentiandionei
dependent).iOnitheiotherihandiANOVAiisiusediwheniweihaveitwoiorimoreithanitwoiindepende
ntivariablesi(treatment).iSupposeiweiwantitoistudyitheieffectsiofithreeidifferentimodelsiofitea
chingionitheiachievementiofistudents.iInithisicaseiweihaveithreeidifferentisamplesitoibeitreate
diusingithreeidifferentitreatments.iSoiANOVAiisitheisuitableitechniqueitoievaluateitheidifferen
ce.
Thisianalysisiprocessidividesitheitotalivariabilityiintoitwoibasicicomponents:i
i) Between-TreatmentiVariancei
Varianceisimplyimeansidifferenceianditoicalculateitheivarianceiisiaiprocessiofimeasuringiho
wibigitheidifferencesiareiforiaisetiofinumbers.iTheibetween-
treatmentivarianceiisimeasuringihowimuchidifferenceiexistsibetweenitheitreatmenticonditi
ons.iIniadditionitoimeasuringidifferencesibetweenitreatments,itheioveralligoaliofiANOVAiisi
toievaluateitheidifferencesibetweenitreatments.iSpecifically,itheipurposeiforitheianalysisiisi
toidistinguishiisitoidistinguishibetweenitwoialternativeiexplanations.
Theidifferencesibetweenitheitreatmentsihaveibeenicausedibyitheitreatmentieffects.i
Theidifferencesibetweenitheitreatmentsiareisimplyidueitoichance.
Thus,ithereiareialwaysitwoipossibleiexplanationsiforitheivariancei(difference)ithatiexistsibetwe
enitreatmentsi
TreatmentiEffect:iTheidifferencesiareicausedibyitheitreatments.iForitheidataiinitablei8.
1,itheiscoresiinisamplei1iareiobtainediatiroomitemperatureiofi50oiandithatiofisamplei2
iati70oi.iItiisipossibleithatitheidifferenceibetweenisampleiisicausedibyitheidifferenceiini
roomitemperature.i
iChance:iTheidifferencesiareisimplyidueitoichance.iItithereiisinoitreatmentieffect,ieveni
theniweicaniexpectisomeidifferenceibetweenisamples.iTheichanceidifferencesiareiunpl
annediandiunpredictableidifferencesithatiareinoticausedioriexplainedibyianyiactioniofit
7
heiresearcher.iResearchersicommonlyiidentifyitwoiprimaryisourcesiforichanceidifferenc
es.i
iiiiiIndividualiDifferencesiEachiparticipantiofitheistudyihasiitsiowniindividualicharacteri
stics.iAlthoughiitiisireasonableitoiexpectithatidifferentisubjectsiwilliproduceidifferentisc
ores,iitiisiimpossibleitoipredictiexactlyiwhatitheidifferenceiwillibe.i
iiiiiExperimentaliErroriInianyimeasurementithereiisiaichanceiofisomeidegreeiofierror.i
Thus,iifiairesearcherimeasuresitheisameiindividualsitwiceiunderisameiconditions,itherei
isigreateripossibilityitoiobtainitwoidifferentimeasurements.iOftenitheseidifferencesiarei
unplannediandiunpredictable,isoitheyiareiconsidereditoibeibyichance.
Thus,iwheniweicalculateitheibetween-
treatmentivariance,iweiareimeasuringidifferencesithaticouldibeieitheribyitreatmentieffectioric
ouldisimplyibeidueitoichance.iIniorderitoidemonstrateithatitheidifferenceiisireallyiaitreatmenti
effect,iweimustiestablishithatitheidifferencesibetweenitreatmentsiareibiggerithaniwouldibeiex
pectedibyichanceialone.iToiaccomplishithisigoal,iweiwillidetermineihowibigitheidifferencesiisiw
henithereiisinoitreatmentieffectiinvolved.iThatiis,iweiwillimeasureihowimuchidifferencei(varian
ce)ioccurredibyichance.iToimeasureichanceidifferences,iweicomputeitheivarianceiwithinitreat
ments.
2)iiiiiWithin-Treatment
iVarianceiWithinieachitreatmenticondition,iweihaveiaisetiofiindividualsiwhoiareitreatediexactly
itheisameianditheiresearcheridoesinotidoianythingithatiwouldicauseitheseiindividualiparticipan
tsitoihaveidifferentiscores.iForiexample,iinitablei8.1itheidataishowsithatifiveiindividualsiwereitr
eatediatiai70oiroomitemperature.iAlthough,itheseifiveistudentsiwereiallitreatediexactlyitheisa
me,ithereiscoresiareidifferent.iQuestioniisiwhyiareitheiscoreidifferent?iAiplainiansweriisithatiiti
isidueitoichance,itheioverallianalysisiofivarianceiandiidentifiesitheisourcesiofivariabilityithatiare
imeasuresibyieachiofitwoibasicicomponents.
Types
OneiWayiANOVAi(LogiciandiProcedure)i
8
Theioneiwayianalysisiofivariancei(ANOVA)iisianiextensioniofiindependentitwo-
sampleittest.iItiisiaistatisticalitechniqueibyiwhichiweicanitestiifithreeiorimoreimeansiareiequal.i
Ititestsiifitheivalueiofiaisingleivariableidiffersisignificantlyiamongithreeiorimoreileveliofiaifactor
.iWeicanialsoisayithationeiwayiANOVAiisiaiprocedureiofitestingihypothesisithatiKipopulationim
eansiareiequal,iwhereiKi≥i2.iIticomparesitheimeansiofitheisamplesiorigroupsiiniorderitoimakeii
nferencesiaboutitheipopulationimeans.iSpecifically,iititestsitheinullihypothesis:iHoi:iµ1i=iµ2i=iµ
3i=i...i=iµk
Whereiµi=igroupimeaniandiki=inumberiofigroupsi
IfioneiwayiANOVAiyieldsistatisticallyisignificantiresult,iweiacceptitheialternateihypothesisi(HA),
iwhichistatesithatithereiareitwoigroupimeansithatiareistatisticallyisignificantlyidifferentifromie
achiother.iHereiitishouldibeikeptiinimindithationeiwayiANOVAicannotitelliwhichispecificigroups
iwereistatisticallyisignificantlyidifferentifromieachiother.iToidetermineiwhichispecificigroupsiar
eidifferentifromieachiother,iairesearcheriwillihaveitoiuseipostihocitest.iAsithereiisionlyioneiind
ependentivariableiorifactoriinioneiwayiANOVAisoiitiisialsoicalledisingleifactoriANOVA.iTheiinde
pendentivariableihasinominalilevelsioriaifewiordinalilevels.iAlso,ithereiisionlyioneidependentiv
ariableiandihypothesesiareiformulatediaboutitheimeansiofitheigroupionidependentivariable.iT
heidependentivariableidifferentiatesiindividualsionisomeiquantitativeidimension.
MultipleiComparisoniProcedurei
Inione-wayiANOVAi“R2”imeasuresitheieffectisize,iitisuffersioneipossibleilimitationi–
iitidoesinotiindicateiwhichigroupimayibeitheiresponsibleiforiaisignificantieffect.iAllithatiaisignifi
cantiR2iandiFistatisticisayiisithatitheimeansiforitheigroupsiareiunlikelyitoihaveibeenisampledifr
omiaisingleihatiofimeans.iUnfortunately,ithereiisinoisimple,iunequivocalistatisticalisolutionitoit
heiproblemioficomparingiforidifferentilevelsiofianiANOVAifactor.iAinumberiofistatisticalimetho
dsihaveibeenidevelopeditoitestiforitheidifferenceiinimeansiamongitheilevelsiofianiANOVAifact
or.iCollectivelyitheseiareiknowniasimultipleicomparisoniproceduresi(MCPs)iorisometimes,iasip
ostihoci(i.e.iafteritheifact)itests.iTheseitestsishouldibeiusediregardediasianiafterthoughtithaniai
rigorousiexaminationiofiprespecifiedihypotheses.iMostiofitheimultiple-
comparisonsimethodsiareimeantitoipair-
9
wiseicomparisonsiofigroupimeans,itoidetermineiwhichiareisignificantlyifromiwhichiothers.iThei
mainipurposeiofimostimultiple-
comparisoniproceduresiisitoicontrolitheioverallisignificanceilevel,iforisomeisetiofiinterferencesi
performediasiaifollow-
upitoiANOVA.iThisioverallisignificanceileveliisitheiprobability,iconditionalioniallitheinullihypoth
esesibeingitestedibeingitrue,iofirejectingiatileastioneiofithem,ioriequivalently,iofihavingiatileas
tioneiconfidenceiintervalinotiincludeitheitrueivalue.iTheivariousimethodsidifferiinihowiwellithe
yiproperlyicontrolitheioverallisignificanceileveliandiinitheirirelativeipower.iCommonlyiusedimet
hodisanditheirirelativeipoweriisigivenibelow.i
iBonferronii–iItiisiextremelyigeneraliandisimple,ibutiofteninotipowerful.i
iTucky’si–iItiisitheibestiofiallipossibleipair-
wiseicomparisonsiwhenisampleisizesiareiunequalioriconfidenceiintervalsiareineeded.iItiisialsoiv
eryigoodieveniwithiequalisampleisizesiwithouticonfidenceiintervals.
iiStepdowni–iItiisitheimostipowerfuliforiallipossibleipair-
wiseicomparisonsiwhenisampleisizesiareiequal.i
iDunnett’si–
iItiisisuitableiforicomparingioneisampleitoieachiofitheiothers,ibutinoticomparingitheiothersitoi
eachiother.i
iHsu’siMCBi–iIticomparesieachimeanitoitheibestiofitheiotherimeans.i
iScheffè’si–iItiisisuitableiforiunplannedicontrastsiamongisetsiofimeans.
iiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiii-------------------------------------
Question.no:03:-
Whatiisitheirangeioficorrelationicoefficient?iExplainistrong,imoderateiandiweakirelationship.
Answer:
10
Correlationicoefficientsiareiindicatorsiofitheistrengthiofitheilinearirelationshipibetweenitwoidif
ferentivariables,ixiandiy.iAilinearicorrelationicoefficientithatiisigreaterithanizeroiindicatesiaipos
itiveirelationship.iAivalueithatiisilessithanizeroisignifiesiainegativeirelationship.iFinally,iaivaluei
ofizeroiindicatesinoirelationshipibetweenitheitwoivariablesixiandiy.iThisiarticleiexplainsitheisig
nificanceiofilinearicorrelationicoefficientiforiinvestors,ihowitoicalculateicovarianceiforistocks,ia
ndihowiinvestorsicaniuseicorrelationitoipredictitheimarket.
UnderstandingiCorrelation
Theicorrelationicoefficienti(ρ)iisiaimeasureithatideterminesitheidegreeitoiwhichitheimovement
iofitwoidifferentivariablesiisiassociated.iTheimosticommonicorrelationicoefficient,igeneratedib
yitheiPearsoniproduct-
momenticorrelation,iisiuseditoimeasureitheilinearirelationshipibetweenitwoivariables.iHoweve
r,iiniainon-
linearirelationship,ithisicorrelationicoefficientimayinotialwaysibeiaisuitableimeasureiofidepend
ence.
Theipossibleirangeiofivaluesiforitheicorrelationicoefficientiisi-
1.0itoi1.0.iIniotheriwords,itheivaluesicannotiexceedi1.0ioribeilessithani-1.0.iAicorrelationiofi-
1.0iindicatesiaiperfectinegativeicorrelation,iandiaicorrelationiofi1.0iindicatesiaiperfectipositivei
correlation.iIfitheicorrelationicoefficientiisigreaterithanizero,iitiisiaipositiveirelationship.iConver
sely,iifitheivalueiisilessithanizero,iitiisiainegativeirelationship.iAivalueiofizeroiindicatesithatither
eiisinoirelationshipibetweenitheitwoivariables.
CorrelationianditheiFinancialiMarkets
Initheifinancialimarkets,itheicorrelationicoefficientiisiuseditoimeasureitheicorrelationibetweeni
twoisecurities.iForiexample,iwhenitwoistocksimoveiinitheisameidirection,itheicorrelationicoeffi
cientiisipositive.iConversely,iwhenitwoistocksimoveiinioppositeidirections,itheicorrelationicoeff
icientiisinegative.
Ifitheicorrelationicoefficientiofitwoivariablesiisizero,ithereiisinoilinearirelationshipibetweenithe
ivariables.iHowever,ithisiisionlyiforiailinearirelationship.iItiisipossibleithatitheivariablesihaveiais
11
trongicurvilinearirelationship.iWhenitheivalueiofiρiisicloseitoizero,igenerallyibetweeni-
0.1iandi+0.1,itheivariablesiareisaiditoihaveinoilinearirelationshipi(oriaiveryiweakilinearirelation
ship).
Foriexample,isupposeithatitheipricesioficoffeeiandicomputersiareiobservediandifounditoihavei
aicorrelationiofi+.0008.iThisimeansithatithereiisinoicorrelation,iorirelationship,ibetweenitheitw
oivariables.
Calculatingiρ
Theicovarianceiofitheitwoivariablesiiniquestionimustibeicalculatedibeforeitheicorrelationicanib
eidetermined.iNext,ieachivariable'sistandardideviationiisirequired.iTheicorrelationicoefficientiis
ideterminedibyidividingitheicovarianceibyitheiproductiofitheitwoivariables'istandardideviations
.
Standardideviationiisiaimeasureiofitheidispersioniofidataifromiitsiaverage.iCovarianceiisiaimeas
ureiofihowitwoivariablesichangeitogether.iHowever,iitsimagnitudeiisiunbounded,isoiitiisidifficu
ltitoiinterpret.iTheinormalizediversioniofitheistatisticiisicalculatedibyidividingicovarianceibyithe
iproductiofitheitwoistandardideviations.iThisiisitheicorrelationicoefficient.
PositiveiCorrelation/strong
Aipositiveicorrelation—whenitheicorrelationicoefficientiisigreaterithani0—
signifiesithatibothivariablesimoveiinitheisameidirection.iWheniρiisi+1,iitisignifiesithatitheitwoiv
ariablesibeingicomparedihaveiaiperfectipositiveirelationship;iwhenioneivariableimovesihigheri
orilower,itheiotherivariableimovesiinitheisameidirectioniwithitheisameimagnitude.
Theicloseritheivalueiofiρiisitoi+1,itheistrongeritheilinearirelationship.iForiexample,isupposeithe
ivalueiofioilipricesiisidirectlyirelateditoitheipricesiofiairplaneitickets,iwithiaicorrelationicoefficie
ntiofi+0.95.iTheirelationshipibetweenioilipricesiandiairfaresihasiaiveryistrongipositiveicorrelatio
nisinceitheivalueiisicloseitoi+1.iSo,iifitheipriceiofioilidecreases,iairfaresialsoidecrease,iandiifithe
ipriceiofioiliincreases,isoidoitheipricesiofiairplaneitickets.
12
Understandingitheicorrelationibetweenitwoistocksi(oriaisingleistock)iandiitsiindustryicanihelpii
nvestorsigaugeihowitheistockiisitradingirelativeitoiitsipeers.iAllitypesiofisecurities,iincludingibo
nds,isectors,iandiETFs,icanibeicomparediwithitheicorrelationicoefficient.i
NegativeiCorrelation
Ainegativei(inverse)icorrelationioccursiwhenitheicorrelationicoefficientiisilessithani0.iThisiisiani
indicationithatibothivariablesimoveiinitheioppositeidirection.iInishort,ianyireadingibetweeni0ia
ndi-1imeansithatitheitwoisecuritiesimoveiinioppositeidirections.iWheniρiisi-
1,itheirelationshipiisisaiditoibeiperfectlyinegativelyicorrelated.iInishort,iifioneivariableiincrease
s,itheiotherivariableidecreasesiwithitheisameimagnitudei(andiviceiversa).iHowever,itheidegreei
toiwhichitwoisecuritiesiareinegativelyicorrelatedimightivaryioveritimei(anditheyiareialmostinev
eriexactlyicorrelatediallitheitime).i
ExamplesiofiNegativeiCorrelation
Foriexample,isupposeiaistudyiisiconducteditoiassessitheirelationshipibetweenioutsideitempera
tureiandiheatingibills.iTheistudyiconcludesithatithereiisiainegativeicorrelationibetweenitheipric
esiofiheatingibillsianditheioutdooritemperature.iTheicorrelationicoefficientiisicalculateditoibei-
0.96.iThisistronginegativeicorrelationisignifiesithatiasitheitemperatureidecreasesioutside,itheip
ricesiofiheatingibillsiincreasei(andiviceiversa).
Wheniiticomesitoiinvesting,iainegativeicorrelationidoesinotinecessarilyimeanithatitheisecuritie
sishouldibeiavoided.iTheicorrelationicoefficienticanihelpiinvestorsidiversifyitheiriportfolioibyiin
cludingiaimixiofiinvestmentsithatihaveiainegative,iorilow,icorrelationitoitheistockimarket.iInish
ort,iwhenireducingivolatilityiriskiiniaiportfolio,isometimesioppositesidoiattract.ii
Foriexample,iassumeiyouihaveiai$100,000ibalancediportfolioithatiisiinvestedi60%iinistocksiand
i40%iinibonds.iIniaiyeariofistrongieconomiciperformance,itheistockicomponentiofiyouriportfoli
oimightigenerateiaireturniofi12%iwhileitheibondicomponentimayireturni-
2%ibecauseiinterestiratesiareirisingi(whichimeansithatibondipricesiareifalling).iThus,itheioverall
ireturnioniyouriportfolioiwouldibei6.4%i((12%ixi0.6)i+i(-
2%ixi0.4).iTheifollowingiyear,iasitheieconomyislowsimarkedlyiandiinterestiratesiareilowered,iy
13
ouristockiportfolioimightigeneratei-
5%iwhileiyouribondiportfolioimayireturni8%,igivingiyouianioveralliportfolioireturniofi0.2%.
Whatiif,iinsteadiofiaibalancediportfolio,iyouriportfolioiwasi100%iequities?iUsingitheisameiretu
rniassumptions,iyouriall-equityiportfolioiwouldihaveiaireturniofi12%iinitheifirstiyeariandi-
5%iinitheisecondiyear.iTheseifiguresiareiclearlyimoreivolatileithanitheibalancediportfolio'siretu
rnsiofi6.4%iandi0.2%.
LineariCorrelationiCoefficient
Theilinearicorrelationicoefficientiisiainumbericalculatedifromigivenidataithatimeasuresitheistre
ngthiofitheilinearirelationshipibetweenitwoivariables,ixiandiy.iTheisigniofitheilinearicorrelationi
coefficientiindicatesitheidirectioniofitheilinearirelationshipibetweenixiandiy.iWheniri(theicorrel
ationicoefficient)iisineari1iori−1,itheilinearirelationshipiisistrong;iwheniitiisineari0,itheilinearirel
ationshipiisiweak.
Eveniforismallidatasets,itheicomputationsiforitheilinearicorrelationicoefficienticanibeitooilongit
oidoimanually.iThus,idataiareioftenipluggediintoiaicalculatorior,imoreilikely,iaicomputerioristat
isticsiprogramitoifinditheicoefficient.
iiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiii----------------------------------
Question.no:04i:-
Explainichiisquareiindependenceitest.iIniwhatisituationishouldiitibeiapplied?
Answer:
WhatiisiaiChi-SquareiStatistic?i
AiChi-
SquareiStatisticiisioneiwayitoiairelationshipibetweenitwoicategoricali(nonnumerical)ivariables.i
TheiChi-
SquareiStatisticiisiaiisiaisingleinumberithatitellsiusihowimuchidifferenceiexistsibetweenitheiobs
14
ervedicountsianditheicountsithationeiexpectsiifithereiisinoirelationshipiinitheipopulation.iTher
eiareitwoidifferentitypesiofichi-squareitests,ibothiinvolveicategoricalidata.iTheseiare:
ia)iAichi-squareigoodnessiofifititest,iandi
b)iAichi-
squareitestiofiindependence.iInitheicomingilinesitheseitestsiwillibeidealtiinisomeidetails.
Theichi-squarei(χ2i)igoodnessiofifititesti(commonlyireferreditoiasione-sampleichi-
square)iisitheimosticommonlyiusedigoodnessiofifititest.iItiexploresitheiproportionioficasesithat
ifalliintoitheivariousicategoriesiofiaisingleivariable,iandicomparesitheseiwithihypothesizedivalu
es.iInisomeisimpleiwordsiweicanisayithatiitiisiuseditoifindioutihowitheiobservedivalueiofiaigive
niphenomenaiisisignificantlyidifferentifromitheiexpectedivalue.iOriweicanialsoisayithatiitiisiuse
ditoitestiifisampleidataifitsiaidistributionifromiaicertainipopulation.iIniotheriwordsiweicanisayit
hatichi-
squareigoodnessiofifititestitellsiusiifitheisampleidatairepresentsitheidataiweiexpectitoifindiinith
eiactualipopulation.iItitellsiusiwhetherisampleidataiareiconsistentiwithiaihypothesizedidistribut
ion.iThisiisiaivariationiofimoreigeneralichi-
squareitest.iTheisettingiforithisitestiisiaisingleicategoricalivariableithaticanihaveimanyilevels.iIni
chi-
squareigoodnessiofifititestisampleidataiisidividediintoiintervals.iThen,itheinumbersiofipointsith
atifalliintoitheiintervalsiareicomparediwithitheiexpectedinumbersiofipointsiinieachiinterval.i.iT
heinullihypothesisiforitheichi-
squareigoodnessiofifititestiisithatitheidataidoesinoticomeifromitheispecifiedidistribution.iTheial
ternateihypothesisiisithatitheidataicomesifromitheispecifiedidistribution.iTheiformulaiforichi-
squareigoodnessiofifititestiis:i
χi2i=i∑ii(observedivaluesi–iexpectedivalues)^2/expectedivalues
Foriusingichi-
squarei(χ2i)igoodnessiofifititestiweiwillihaveitoisetiupinulliandialternateihypothesis.iAinullihyp
15
othesisiassumesithatithereiisinoisignificanceidifferenceibetweeniobservediandiexpectedivalue.i
Then,ialternateihypothesisiwillibecome,ithereiisisignificantidifferentidifferenceibetweenitheiob
servedianditheiexpectedivalue.iNowicomputeitheivalueiofichi-squareiofifititestiusingiformula:
χi2i=i∑ii(observedivaluesi–iexpectedivalues)^2/expectedivalues
Twoipotentialidisadvantagesiofichi-squareiare:
a)iTheichi-
squareitesticanionlyibeiuseditoiputidataiintoiclasses.iIfithereiisidataithatihaveinotibeeniputiint
oiclassesitheniitiisinecessaryitoimakeiaifrequencyitableiofihistogramibeforeiperformingitheitest
.i
b)iItirequiresisufficientisampleisizeiiniorderiforichi-squareiapproximationitoibeivalid
Chi-SquareiIndependenceiTesti
Aichi-squarei(χ2i)itestiofiindependenceiisitheisecondiimportantiformiofichi-
squareitests.iItiisiuseditoiexploreitheirelationshipibetweenitwoicategoricalivariables.iEachiofith
eseivariablesicanihaveitwoiofimoreicategories.iItideterminesiifithereiisiaisignificantirelationship
ibetweenitwoinominali(categorical)ivariables.iTheifrequencyiofioneinominalivariableiisicompar
ediwithidifferentivaluesiofitheisecondinominalivariable.iTheidataicanibeidisplayediiniR*Ciconti
ngencyitable,iwhereiRiisitheirowiandiCiisitheicolumn.iForiexample,itheiresearcheriwantsitoiexa
mineitheirelationshipibetweenigenderi(maleiandifemale)iandiempathyi(highivs.ilow).iTheiresea
rcheriwilliuseichi-
squareitestiofiindependence.iIfitheinullihypothesisiisiacceptedithereiwouldibeinoirelationshipib
etweenigenderiandiempathy.iIfitheinullihypothesisiisirejectedithenitheiconclusioniwillibeithere
iisiairelationshipibetweenigenderiandiempathyi(e.g.isayifemalesitentitoiscoreihigherioniempat
hyiandimalesitenditoiscoreilowerioniempathy).iTheichi-
squareitestiofiindependenceibeingiainon-
parametricitechniqueifollowilessistrictiassumptions,ithereiareisomeigeneraliassumptionsiwhich
ishouldibeitakenicareiof:
16
i) RandomiSamplei-iSampleishouldibeiselectediusingisimpleirandomisamplingimethod.i
ii) Variablesi-iBothivariablesiunderistudyishouldibeicategorical.i
iii) IndependentiObservationsi–
iEachipersonioricaseishouldibeicountedionlyionceiandinoneishouldiappeariinimoreitha
nioneicategoryiofigroup.iTheidataifromioneisubjectishouldinotiinfluenceitheidataifromi
anotherisubject.i
iv) Ifitheidataiareidisplayediiniaicontingencyitable,itheiexpectedifrequencyicountiforieachi
celliofitheitableiisiatileasti5.iBothitheichi-
squareitestsiareisometimeiconfusedibutitheyiareiquiteidifferentifromieachiother.i
iTheichi-
squareitestiforiindependenceicomparesitwoisetsiofidataitoiseeiifithereiisirelationship.i
iTheichi-squareigoodnessiofifititestiisitoifitioneicategoricalivariableitoiaidistribution.
iiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiii----------------------------------
Question.no:05:-
CorrelationiisipreirequisiteiofiRegressioniAnalysis.iExplain.
Answer:
Correlationiisiaistatisticalitechniqueiuseditoimeasureiandidescribeirelationshipibetweenitwoiva
riables.iTheseivariablesiareineitherimanipulatedinoricontrolled,iratheritheyisimplyiareiobserve
diasitheyinaturallyiexistiinitheienvironment.iSupposeiairesearcheriisiinterestediinirelationshipib
etweeninumberiofichildreniiniaifamilyiandiIQiofitheiindividualichild.iHeiwoulditakeiaigroupiofis
tudentsicomingifromidifferentifamilies.iTheniheisimplyiobserveiorirecorditheinumberiofichildre
niiniaifamilyiandithenimeasureiIQiscoreiofieachiindividualistudentisameigroup.iHeiwillineitheri
manipulateinoricontrolianyivariable.iCorrelationirequiresitwoiseparateiscoresiforieachiindividu
ali(oneiscoreifromieachiofitwoivariables).iTheseiscoresiareinormallyiidentifiediasiXiandiYiandic
anibeipresentediiniaitableioriiniaigraph.
17
Regressioni
Aicorrelationiquantifiesitheidegreeiandidirectionitoiwhichitwoivariablesiareirelated.iItidoesinot
ifitiailineithroughitheidataipoints.iItidoesinotihaveitoithinkiaboutitheicauseiandieffect.iItidoesin
otinatteriwhichiofitheitwoivariablesiisicalledidependentiandiwhichiisicallediindependent.iOnith
eiotherihandiregressionifindsitheibestilineithatipredictsidependentivariablesifromitheiindepen
dentivariable.iTheidecisioniofiwhichivariableiisicallsidependentiandiwhichicallsiindependentiisi
aniimportantimatteriiniregression,iasiitiwilligetiaidifferentibest-
fitilineiifiweiexchangeitheitwoivariables,ii.e.idependentitoiindependentiandiindependentitoide
pendent.iTheilineithatibestipredictsiindependentivariableifromidependentivariableiwillinotibeit
heisameiasitheilineithatipredictsidependentivariableifromiindependentivariable.
LetiusistartiwithitheisimpleicaseiofistudyingitheirelationshipibetweenitwoivariablesiXiandiY.iTh
eivariableiYiisidependentivariableianditheivariableiXiisitheiindependentivariable.iWeiareiintere
stediiniseeingihowivariousivaluesiofitheiindependentivariableiXipredicticorrespondingivaluesiof
idependentiY.iThisistatisticalitechniqueiisicallediregressionianalysis.iWeicanisayithatiregressioni
analysisiisiaitechniqueithatiisiuseditoimodelitheidependencyiofioneidependentivariableiuponio
neiindependentivariable.iMerriam-
Websterionlineidictionaryidefinesiregressioniasiaifunctionalirelationshipibetweenitwoiorimorei
correlatedivariablesithatiisiofteniempiricallyideterminedifromidataiandiisiusediespeciallyitoipre
dictivaluesiofioneivariableiwhenigivenivariablesiofiothers.iAccordingitoiGravetteri&iWallnuai(2
002),iregressioniisiaistatisticalitechniqueiforifindingitheibest-
fittingistraightilineiforiaisetiofidataiisicallediregression,ianditheiresultingistraightilineiisicalledir
egressioniline.
ObjectivesiofiRegressioniAnalysisi
Theiregressionianalysisiisiuseditoiexplainivariabilityiinidependentivariableibyimeaniofioneiorim
oreiofiindependentivariablesianditoianalyzeirelationshipsiamongivariablesitoiansweritheiquesti
oniofihowimuchidependentivariableichangesiwithitheichangesiinitheiindependentivariablesian
ditoiforecastioripredictitheivalueiofidependentivariableibasedionitheivaluesiofitheiindependent
18
ivariable.iTheiprimaryiobjectiveiofitheiregressioniisitoidevelopiairelationshipibetweeniairespon
seivariableianditheiexplanatoryivariableiforitheipurposeiofiprediction,iassumesithatiaifunctiona
lirelationshipiexists,iandialternativeiapproachesiareisuperior
WhyidoiweiuseiRegressioniAnalysis?i
Regressionianalysisiestimatesitheirelationshipibetweenitwoiorimoreivariablesiandiisiusediforifo
recastingiorifindingicauseiandieffectirelationshipibetweenitheivariables.iThereiareimultipleiben
efitsiofiusingiregressionianalysis.iTheseiareiasifollows:ii)iItiindicatesitheisignificantirelationships
ibetweenidependentianditheiindependentivariables.iii)iItiindicatesitheistrengthiofiimpactiofim
ultipleiindependentivariablesioniaidependentivariable.iiii)iItiallowsiusitoicompareitheieffectsiof
ivariablesimeasuredionidifferentiscales.iTheseibenefitsihelpiairesearcheritoiestimateiandievalu
ateitheibestisetiofivariablesitoibeiusediforibuildingiproductiveimodels.
TypesiofiRegressioni
Commonlyiuseditypesiofiregressioniare:i
LineariRegression:iItiisitheimosticommonlyiuseditypesiofiregression.iInithisitechniqueit
heidependentivariableiisicontinuousianditheiindependentivariableicanibeicontinuousior
idiscreteianditheinatureiofiregressionilineiisilinear.iLineariregressioniestablishesiairelati
onshipibetweenidependentivariablei(Y)iandioneiorimoreiindependentivariablesi(X)iusin
gibestifitistraightilinei(alsoiknowniasiregressioniline).
LogisticiRegression:iLogisticiregressioniisiaistatisticalimethodiforianalyzingiaidatasetiini
whichithereiareioneiorimoreiindependentivariablesithatidetermineianioutcome.iTheiou
tcomeiisimeasurediwithitheidichotomousi(binary)ivariable.iLikeialliregressionianalysis,it
heilogisticiregressioniisiaipredictiveianalysis.iItiisiuseditoidescribeiandiexplainirelations
hipibetweenioneidependentibinaryivariableiandioneiorimoreinominal,iordinal,iintervali
oriratioileveliindependentivariables.i
PolynomialiRegression:iItiisiaiformiofiregressionianalysisiiniwhichitheirelationshipibetw
eeniindependentivariableiXiandidependentivariableiYiisimodelediasianinthidegreeipoly
19
nomialiinix.ithisitypeiofiregressionifitsiainon-
linearirelationshipibetweenitheivaluesiofiXiwithitheicorrespondingivaluesiofiY.i
StepwiseiRegression:iItiisiaimethodiofifittingiregressionimodeliiniwhichitheichoiceiofip
redictiveivariablesiisicarriedioutibyianiautomaticiprocedure.iInieachistep,iaivariableiisic
onsiderediforiadditioniorisubtractionifromitheisetiofiexplanatoryivariablesibasedioniso
meipre-
specifiedicriteria.iTheigeneraliideaibehindithisiprocedureiisithatiweibuildiouriregression
imodelifromiaisetiofipredictorivariableibyienteringiandiremovingipredictorsiiniourimod
el,iiniaistepwiseimanner,iuntilithereiisinoijustifiableireasonitoienterioriremoveianyimor
e.i
RidgeiRegressioni:Itiisiaitechniqueiforianalyzingimultipleiregressionidataithatisufferifro
mimulticollinearityi(independentivariablesiareihighlyicorrelated).iWhenimulticollinearit
yioccurs,ileastisquaresiestimatesiareiunbiased,ibutitheirivariancesiareilargeisoithatithey
imayibeifarifromitheitrueivalue.iByiaddingitheidegreeiofibiasitoitheiregressioniestimate
s,iridgeiregressionireducesitheistandardierrors.i
LASSOiRegression:iLASSOiorilassoistandsiforiLeastiAbsoluteiShrinkageiandiSelectioniOp
erator.iItiisiaimethodithatiperformsibothivariableiselectioniandiregularizationiiniorderit
oienhanceitheipredictioniaccuracyiandiinterpretabilityiofitheistatisticalimodeliitiproduc
es.iThisitypeiofiregressioniusesishrinkage.iShrinkageiisiwhereidataivaluesiareishrunkito
wardsiaicentralipoint,ilikeitheimean.i
ElasticiNetiRegression:iThisitypeiofiregressioniisiaihybridiofilassoiandiridgeiregressionit
echniques.iItiisiusefuliwhenithereiareimultipleifeaturesiwhichiareicorrelated.
______________________iiiiiiiiiiii