概率论毕业论文外文翻译.docx
《概率论毕业论文外文翻译.docx》由会员分享,可在线阅读,更多相关《概率论毕业论文外文翻译.docx(15页珍藏版)》请在冰点文库上搜索。
概率论毕业论文外文翻译
[]
济南大学泉城学院毕业论文外文资料翻译
Statisticalhypothesistesting
AdrianaAlbu,LoredanaUngureanu
PolitehnicaUniversityTimisoara,adrianaa@aut.utt.ro
PolitehnicaUniversityTimisoara,loredanau@aut.utt.ro
AbstractInthisarticle,wepresentaBayesianstatisticalhypothesistestinginspection,testingtheoryandtheprocessMentionedhypothesistestingintherealworldandtheimportanceof,andsuccessfultestoftheNotes.
KeywordsBayesianhypothesistesting;Bayesianinference;TestofsignificanceIntroduction
Astatisticalhypothesistestisamethodofmakingdecisionsusingdata,whetherfromacontrolledexperimentoranobservationalstudy(notcontrolled).Instatistics,aresultiscalledstatisticallysignificantifitisunlikelytohaveoccurredbychancealone,accordingtoapre-determinedthresholdprobability,thesignificancelevel.Thephrase"testofsignificance"wascoinedbyRonaldFisher:
"Criticaltestsofthiskindmaybecalledtestsofsignificance,andwhensuchtestsareavailablewemaydiscoverwhetherasecond
sampleisorisnotsignificantlydifferentfromthefirst."
[1]
Hypothesistestingissometimescalledconfirmatorydataanalysis,incontrasttoexploratorydataanalysis.Infrequencyprobability,thesedecisionsarealmostalwaysmadeusingnull-hypothesistests.TheseareteststhatanswerthequestionAssumingthatthenullhypothesisistrue,whatistheprobabilityofobservingavaluefortheteststatisticthatisatleastasextremeasthevaluethatwasactuallyobserved?
)2Moreformally,theyrepresentanswerstothequestion,posedbeforeundertakinganexperiment,ofwhatoutcomesoftheexperimentwouldleadtorejectionofthenullhypothesisforapre-specifiedprobabilityofanincorrectrejection.Oneuseofhypothesistestingisdecidingwhetherexperimentalresultscontainenoughinformationtocastdoubtonconventionalwisdom.
Statisticalhypothesistestingisakeytechniqueoffrequentiststatisticalinference.TheBayesianapproachtohypothesistestingistobaserejectionofthehypothesisonthe
posteriorprobability.
[3][4]
Otherapproachestoreachingadecisionbasedondataare
availableviadecisiontheoryandoptimaldecisions.
Thecriticalregionofahypothesistestisthesetofalloutcomeswhichcausethenullhypothesistoberejectedinfavorofthealternativehypothesis.ThecriticalregionisusuallydenotedbytheletterC.
One-sampletestsareappropriatewhenasampleisbeingcomparedtothepopulationfromahypothesis.Thepopulationcharacteristicsareknownfromtheoryorarecalculatedfromthepopulation.
-1-
济南大学泉城学院毕业论文外文资料翻译
Two-sampletestsareappropriateforcomparingtwosamples,typicallyexperimentalandcontrolsamplesfromascientificallycontrolledexperiment.
Pairedtestsareappropriateforcomparingtwosampleswhereitisimpossibletocontrolimportantvariables.Ratherthancomparingtwosets,membersarepairedbetweensamplessothedifferencebetweenthemembersbecomesthesample.Typicallythemeanofthedifferencesisthencomparedtozero.
Z-testsareappropriateforcomparingmeansunderstringentconditionsregardingnormalityandaknownstandarddeviation.
T-testsareappropriateforcomparingmeansunderrelaxedconditions(lessisassumed).
Testsofproportionsareanalogoustotestsofmeans(the50%proportion).Chi-squaredtestsusethesamecalculationsandthesameprobabilitydistributionfor
differentapplications:
Chi-squaredtestsforvarianceareusedtodeterminewhetheranormalpopulationhasaspecifiedvariance.Thenullhypothesisisthatitdoes.
Chi-squaredtestsofindependenceareusedfordecidingwhethertwovariablesareassociatedorareindependent.Thevariablesarecategoricalratherthannumeric.Itcanbeusedtodecidewhetherleft-handednessiscorrelatedwithlibertarianpolitics(ornot).Thenullhypothesisisthatthevariablesareindependent.Thenumbersusedinthecalculationaretheobservedandexpectedfrequenciesofoccurrence(fromcontingencytables).
Chi-squaredgoodnessoffittestsareusedtodeterminetheadequacyofcurvesfittodata.Thenullhypothesisisthatthecurvefitisadequate.Itiscommontodeterminecurveshapestominimizethemeansquareerror,soitisappropriatethatthegoodness-of-fitcalculationsumsthesquarederrors.
F-tests(analysisofvariance,ANOVA)arecommonlyusedwhendecidingwhethergroupingsofdatabycategoryaremeaningful.Ifthevarianceoftestscoresoftheleft-handedinaclassismuchsmallerthanthevarianceofthewholeclass,thenitmaybeusefultostudyleftiesasagroup.Thenullhypothesisisthattwovariancesarethesame-sotheproposedgroupingisnotmeaningful.
Thetestingprocess
Inthestatisticalliterature,statisticalhypothesistestingplaysafundamentalrole.Theusuallineofreasoningisasfollows:
1.Thereisaninitialresearchhypothesisofwhichthetruthisunknown.
2.Thefirststepistostatetherelevantnullandalternativehypotheses.Thisisimportantasmis-statingthehypotheseswillmuddytherestoftheprocess.Specifically,thenullhypothesisallowsattachinganattribute:
itshouldbechoseninsuchawaythatitallowsustoconcludewhetherthealternativehypothesiscan
eitherbeacceptedorstaysundecidedasitwasbeforethetest.
[9]
3.Thesecondstepistoconsiderthestatisticalassumptionsbeingmadeaboutthesampleindoingthetest;forexample,assumptionsaboutthestatisticalindependenceorabouttheformofthedistributionsoftheobservations.Thisis
-2-
[7]
[8]
济南大学泉城学院毕业论文外文资料翻译
equallyimportantasinvalidassumptionswillmeanthattheresultsofthetestareinvalid.
4.Decidewhichtestisappropriate,andstatetherelevantteststatisticT.
5.Derivethedistributionoftheteststatisticunderthenullhypothesisfromtheassumptions.Instandardcasesthiswillbeawell-knownresult.ForexampletheteststatisticmayfollowaStudent'stdistributionoranormaldistribution.
6.Selectasignificancelevel(α),aprobabilitythresholdbelowwhichthenullhypothesiswillberejected.Commonvaluesare5%and1%.
7.ThedistributionoftheteststatisticunderthenullhypothesispartitionsthepossiblevaluesofTintothoseforwhichthenull-hypothesisisrejected,thesocalledcriticalregion,andthoseforwhichitisnot.Theprobabilityofthecriticalregionisα.
8.Computefromtheobservationstheobservedvaluet
obs
oftheteststatisticT.
9.Decidetoeitherfailtorejectthenullhypothesisorrejectitinfavorofthealternative.ThedecisionruleistorejectthenullhypothesisH0iftheobservedvaluetobsisinthecriticalregion,andtoacceptor"failtoreject"thehypothesisotherwise.
UseandImportance
Statisticsarehelpfulinanalyzingmostcollectionsofdata.Thisisequallytrueofhypothesistestingwhichcanjustifyconclusionsevenwhennoscientifictheoryexists.Realworldapplicationsofhypothesistestinginclude:
Testingwhethermorementhanwomensufferfromnightmares
Establishingauthorshipofdocuments
Evaluatingtheeffectofthefullmoononbehavior
Determiningtherangeatwhichabatcandetectaninsectbyecho
Decidingwhetherhospitalcarpetingresultsinmoreinfections
Selectingthebestmeanstostopsmoking
Checkingwhetherbumperstickersreflectcarownerbehavior
Testingtheclaimsofhandwritinganalysts
Statisticalhypothesistestingplaysanimportantroleinthewholeofstatisticsandinstatisticalinference.Forexample,Lehmann(1992)inareviewofthefundamentalpaperbyNeymanandPearson(1933)says:
"Nevertheless,despitetheirshortcomings,thenewparadigmformulatedinthe1933paper,andthemanydevelopmentscarriedoutwithinitsframeworkcontinuetoplayacentralroleinboththetheoryandpracticeofstatisticsandcanbeexpectedtodosointheforeseeablefuture".
Significancetestinghasbeenthefavoredstatisticaltoolinsomeexperimentalsocialsciences(over90%ofarticlesintheJournalofAppliedPsychologyduringtheearly1990s).Otherfieldshavefavoredtheestimationofparameters.Editorsoftenconsidersignificanceasacriterionforthepublicationofscientificconclusionsbasedonexperimentswithstatisticalresults.
Cautions
Thesuccessfulhypothesistestisassociatedwithaprobabilityandatype-Ierrorrate.Theconclusionmightbewrong.
-3-
[28]
[10]
.
济南大学泉城学院毕业论文外文资料翻译
Theconclusionofthetestisonlyassolidasthesampleuponwhichitisbased.Thedesignoftheexperimentiscritical.Anumberofunexpectedeffectshavebeenobservedincluding:
TheCleverHanseffect.Ahorseappearedtobecapableofdoingsimplearithmetic.TheHawthorneeffect.Industrialworkersweremoreproductiveinbetter
illumination,andmostproductiveinworse.
ThePlaceboeffect.Pillswithnomedicallyactiveingredientswereremarkablyeffective.
Astatisticalanalysisofmisleadingdataproducesmisleadingconclusions.Theissueofdataqualitycanbemoresubtle.Inforecastingforexample,thereisnoagreementonameasureofforecastaccuracy.Intheabsenceofaconsensusmeasurement,nodecisionbasedonmeasurementswillbewithoutcontroversy.
ThebookHowtoLiewithStatisticsisthemostpopularbookonstatisticseverpublished.Itdoesnotmuchconsiderhypothesistesting,butitscautionsareapplicable,including:
Manyclaimsaremadeonthebasisofsamplestoosmalltoconvince.Ifareportdoesnotmentionsamplesize,bedoubtful.
Hypothesistestingactsasafilterofstatisticalconclusions;Onlythoseresultsmeetingaprobabilitythresholdarepublishable.Economicsalsoactsasapublicationfilter;Onlythoseresultsfavorabletotheaut