《计量经济学导论》ppt课件.ppt

《《计量经济学导论》ppt课件.ppt》由会员上传分享，免费在线阅读，更多相关内容在教育资源-天天文库。

1、Chapter7MultipleRegressionAnalysiswithQualitativeInformationWooldridge:IntroductoryEconometrics:AModernApproach,5eQualitativeInformationExamples:gender,race,industry,region,ratinggrade,…AwaytoincorporatequalitativeinformationistousedummyvariablesTheymayappearasthedependentoras

2、independentvariablesAsingledummyindependentvariableDummyvariable:=1ifthepersonisawoman=0ifthepersonisman=thewagegain/lossifthepersonisawomanratherthanaman(holdingotherthingsfixed)MultipleRegressionAnalysis:QualitativeInformationGraphicalIllustrationAlternativeinterpretationofc

3、oefficient:i.e.thedifferenceinmeanwagebetweenmenandwomenwiththesamelevelofeducation.InterceptshiftMultipleRegressionAnalysis:QualitativeInformationDummyvariabletrapThismodelcannotbeestimated(perfectcollinearity)Whenusingdummyvariables,onecategoryalwayshastobeomitted:Alternativ

4、ely,onecouldomittheintercept:ThebasecategoryaremenThebasecategoryarewomenDisadvantages:1)Moredifficulttotestfordiffe-rencesbetweentheparameters2)R-squaredformulaonlyvalidifregressioncontainsinterceptMultipleRegressionAnalysis:QualitativeInformationEstimatedwageequationwithinte

5、rceptshiftDoesthatmeanthatwomenarediscriminatedagainst?Notnecessarily.Beingfemalemaybecorrelatedwithotherproduc-tivitycharacteristicsthathavenotbeencontrolledfor.Holdingeducation,experience,andtenurefixed,womenearn1.81$lessperhourthanmenMultipleRegressionAnalysis:QualitativeIn

6、formationComparingmeansofsubpopulationsdescribedbydummiesDiscussionItcaneasilybetestedwhetherdifferenceinmeansissignificantThewagedifferencebetweenmenandwomenislargerifnootherthingsarecontrolledfor;i.e.partofthedifferenceisduetodiffer-encesineducation,experienceandtenurebetwee

7、nmenandwomenNotholdingotherfactorsconstant,womenearn2.51$perhourlessthanmen,i.e.thedifferencebetweenthemeanwageofmenandthatofwomenis2.51$.MultipleRegressionAnalysis:QualitativeInformationFurtherexample:EffectsoftraininggrantsonhoursoftrainingThisisanexampleofprogramevaluationT

8、reatmentgroup(=grantreceivers)vs.controlgroup(=nogrant)Isthee