《计量经济学导论》ch17ppt课件.ppt

《《计量经济学导论》ch17ppt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、Chapter17LimitedDependentVariableModelsandSampleSelectionCorrectionsWooldridge:IntroductoryEconometrics:AModernApproach,5eLimiteddependentvariables(LDV)LDVarevariableswhoserangeissubstantivelyrestrictedBinaryvariables,e.g.employed/notemployedNonnegativevariables,e.g.wages,prices,inter

2、estratesNonnegativevariableswithexcesszeros,e.g.laborsupplyCountvariables,e.g.thenumberofarrestsinayearCensoredvariables,e.g.unemploymentdurationsSampleselectionmodelsThesampleusedtoinferpopulationrelationshipsisendogenouslyse-lected,e.g.wageofferregressionbutdataonlyaboutworkingwomen

3、LimitedDependentVariablesandSampleSelectionModelsLogitandProbitmodelsforbinaryresponseDisadvantagesoftheLPMforbinarydependentvariablesPredictionssometimesoutsidetheunitintervalPartialeffectsofexplanatoryvariablesareconstantNonlinearmodelsforbinaryresponseResponseprobabilityisanonline

4、arfunctionofexplanat.variablesProbabilityofa„success“givenexplanatoryvariablesAcumulativedistributionfunction..Theresponseprobabilityisthusafunctionoftheexplanatoryvariablesx.Shorthandvectornotation:thevectorofexplanatoryvariablesxalsocontainstheconstantofthemodel.LimitedDependentVari

5、ablesandSampleSelectionModelsChoicesforthelinkfunctionLatentvariableformulationoftheLogitandProbitmodelLogit:(logisticfunction)Probit:(normaldistribution)andIfthelatentvariabley*islargerthanzero,ytakesonthevalue1,ifitislessorequalzero,ytakeson0(y*canthusbeinterpretedasthepropensityto

6、havey=1)LimitedDependentVariablesandSampleSelectionModelsInterpretationofcoefficientsinLogitandProbitmodelsPartialeffectsarenonlinearanddependonthelevelof!whereHowdoestheprobabilityfory=1changeifexplanatoryvariablexjchangesbyoneunit?Continuousexplanatoryvariables:Discreteexplanatoryv

7、ariables:Forexample,explanatoryvariablexkincreasesbyoneunit.LimitedDependentVariablesandSampleSelectionModelsMaximumlikelihoodestimationofLogitandProbitmodelsPropertiesofmaximumlikelihoodestimatorsMaximumlikelihoodestimatorsareconsistent,asymptoticallynormalandasymptoticallyefficient

8、ifthe