第二章 一元線性回歸的檢驗(yàn)和預(yù)測2012 川大 計(jì)量經(jīng)濟(jì)學(xué) 課件

CHAPTERTWOHYPOTHESISTESTINGANDPREDITIONOFTWO-VARIABLELINEARREGRESSIONMODEL1.ttestandconfidenceinterval2.prediction:pointpredictorintervalpredictor3.reportandanalysis4.caseHypothesistest:Isagivenobservationorfindingsufficientlyclosedtohypothesizedvaluesothatwecan’trejectthestatedhypothesis?There’retwoapproachestoproceedthehypothesistest:testofsignificanceandconfidenceinterval2.1testofsignificanceandconfidenceinterval2.1.1testofsignificanceConstitutingstandardizednormaldistributionButtoobtain,weneedUsually,trueisunknown,nowwesubstitutefor.Furthermorewesubstituteforandgetat-statistic1、Putforwardhypothesis:Nullhypothesis:H0：Alternativehypothesis:H1：2、Constitutet-statisticandcomputethevalueofT:3、Decisionrule:If|T|>ta/2(n-2)，rejecttheH0,,itmeansthattheexplanatoryvariableindeedinfectsdependentvariable.ThevariableXisstatisticallysignificant.Otherwise,acceptH0，itindicatesthatthereisn’tsignificantcorrelationbetweenexplanatoryanddependentvariablesupplement：decisionruleoft-testH0:H1:rejectH0雙尾β1=β1*β1≠β1*|T|>tα/2右尾β1≤β1*β1＞β1*T>tα左尾β1≥β1*β1

＜β1*T<-tαα:levelofsignificance,fixed(1-α)%:confidencedf:degreeoffreedomCriticalvalueP-value:probabilityactualprobabilityofobtainingavalueofT-TESTasmuchasorgreaterthanobtainedinOLSestimationIffixedαisgreaterthanP,wemayrejectthenullhypothesis.2.1.2confidenceintervalForstatisticTfollowstdistribution：Givenαlevelofsignificance：Iscalledtheconfidenceintervalofthetrue?givenconfidenceorthelevelofsignificanceIfzerofallswithintheconfidenceinterval,weacceptthenullhypothesis.Otherwise,rejectnullhypothesis.Correspondingly,2.2theanalysisofvarianceSSD.F.ESS1RSSN-2TSSN-12.3predictionMeanpredictionIndividualpredictionWehavesampleregressionfunctionAfuturevalueofXisknownasX0，thenthefutureconditionalmeansofYcanbepredictedasfollowing:IsestimatorofthefutureconditionalmeansofYItiscalledpointpredictorAnotherquestionisintervalpredictor:whichistheconfidenceintervalofgivenconfidence2.3.1MEANPREDICTIONPredictionofconditionalmeansYongivenXαisthelevelofsignificance2.3.2individualpredictionPredictionofanindividualYvalueongivenX1、Thedistributionofresidual2.ConstitutetheT-statisticαisthelevelofsignificanceThepredictiveabilityfallsmarkedlyasX0departsprogressivelyfromthesamplemean!ConfidenceintervalformeansConfidenceintervalforindividualsConclusion:Thegreatern,themoreaccuratepredictionThecloserforXtosamplemeansX0,themoreaccuratepredictionThewidthofconfidencebandsissmallestwhenThewiderthesamplerangeofX,themoreaccurateprediction2.4normalitytestsThehypothesistestisbasedontheassumptionthatthedisturbancefollowsthenormaldistribution. But,infact,isthisassumptionsatisfied?WeusetheresidualtocheckitHistogramofResidualsNormalProbabilityPlot2.5reportandanalysisoftheresultofregression

Analysis:1、Arethesigns

ofestimatedregressioncoefficientaccordancewithourexpectationoreconomictheory?2、Istheslopecoefficientstatisticalsignificant?(statisticallysignificantlydifferentfromZero.)3、HowwelldoestheregressionmodelexplainthechangeorvariationofY?IfR2isgreatenough,wecansaythattheregressionmodelseemsquitegood.4、Isourmodelsatisfiedthenormalityassumption

？案例11977-1991，美國，討論股票和黃金的保值作用，《貨幣戰(zhàn)爭》中的觀點(diǎn)是否得到支持？

每盎司紐約黃金美元價格消費(fèi)價格指數(shù)（CPI）紐約股票交易所指數(shù)(NYSE)147.9860.653.69193.4465.253.7307.6272.658.32612.5182.468.1459.6190.974.02376.0196.568.93423.83