第6章-多重共线性.ppt

下载文档

5
0
约5.63千字
约 18页
2023-12-06 发布于四川
举报
版权申诉
保障服务

第6章-多重共线性.ppt

1、本文档共18页，可阅读全部内容。
2、原创力文档（book118）网站文档一经付费（服务费），不意味着购买了该文档的版权，仅供个人/单位学习、研究之用，不得用于商业用途，未经授权，严禁复制、发行、汇编、翻译或者网络传播等，侵权必究。
3、本站所有内容均由合作方或网友上传，本站不对文档的完整性、权威性及其观点立场正确性做任何保证或承诺！文档内容仅供研究参考，付费前请自行鉴别。如您付费，意味着您自己接受本站规则且自行承担风险，本站不退款、不进行额外附加服务；查看《如何避免下载的几个坑》。如果您已付费下载过本站文档，您可以点击这里二次下载。
4、如文档侵犯商业秘密、侵犯著作权、侵犯人身权等，请点击“版权申诉”（推荐），也可以打举报电话：400-050-0827(电话支持时间：9:00-18:30)。

CHAPTER6MULTICOLLINEARITY

theassumption6ofmultipleregressioncan’tbesatisfied,namely,amongtheexplanatoryvariablesthereismoreorlesslinearfunctionrelatione.g.ThedemandfunctionofchickenY=B0+B1X1i+B2X2i+B3X3i+B4X4iYi-demandofchicken,X1-PDI。X2-thepriceofchicken，X3，X4-priceofporkandbeef5.1thenatureofmulticollinearity

threevariablelinearregressionmodelYt=?0+?1X1t+?2X2t+utE(Yt)=?0+?1X1t+?2X2tsupposethatX1，X2areperfectlylinearlycorrelated,thereforetheircoefficientofcorrelationis1

seetheestimatorofparametersofregressionBecausedenominatoriszero,theregressioncoefficientsareindeterminateandstandarderrorareinfinite

IFmulticollinearityislessthanperfect,theregressioncoefficientisdeterminatebyOLS,butpossesslargestandarderror,namelywithlessprecisionandaccuracySomulticollinearityisaproblemofextent.Inthischapter,wediscusstheimperfectbuthighmulticollinearity.

5.2thesourceandpracticalconsequenceofmulticollinearity5.2.1thesourceofmulticollinearityMulticollinearitymaybeduetothefollowingfactors1.Dataofregressorsshareacommontimetrend.2．Theinherentrelationshipamongtheeconomicvariableespeciallyincross-sectiondata3．Modelspecification,e.g.polynomialmodel,orasfollowingCt=?0+?1Yt+?2Yt-1+?3Yt-2+ut

theoretically,OLSestimatorisstillBLUE，butpractically：1．EventhoughthevarianceofOLSestimatorisminimum,theestimatorshavelargevariance,makingtheestimationdifficult2．Theconfidenceintervalbecomeswiderandtratiobecomessmaller,leadingtotheacceptanceofthe“zeronullhypothesis”andsomeregressorsarestatisticallyinsignificant3.R2,theoverallmeasureofgoodnessoffit,canbeveryhigh4．Theolsestimatorandtheirstandarderrorcanbesensitivetosmallchangesinthedata。5．Thesignofregressioncoefficientmaybewron