欢迎来到天天文库
浏览记录
ID:5534497
大小:1.51 MB
页数:54页
时间:2017-12-17
《线性回归模型在sas_em中的应用实例》由会员上传分享,免费在线阅读,更多相关内容在学术论文-天天文库。
1、Error!Notextofspecifiedstyleindocument.Error!Notextofspecifiedstyleindocument.3-53Chapter3PredictiveModelingUsingRegression3.1IntroductiontoRegression3-33.2RegressioninEnterpriseMiner3-8Error!Notextofspecifiedstyleindocument.Error!Notextofspecifiedstyleindocument.3-53Chapter3Pr
2、edictiveModelingUsingRegression3.1IntroductiontoRegression3-33.2RegressioninEnterpriseMiner3-8Error!Notextofspecifiedstyleindocument.Error!Notextofspecifiedstyleindocument.3-533.1IntroductiontoRegressionError!Notextofspecifiedstyleindocument.Error!Notextofspecifiedstyleindocume
3、nt.3-53TheRegressionnodeinEnterpriseMinerdoeseitherlinearorlogisticregressiondependinguponthemeasurementlevelofthetargetvariable.Linearregressionisdoneifthetargetvariableisanintervalvariable.Inlinearregressionthemodelpredictsthemeanofthetargetvariableatthegivenvaluesoftheinputv
4、ariables.Logisticregressionisdoneifthetargetvariableisadiscretevariable.Inlogisticregressionthemodelpredictstheprobabilityofaparticularlevel(s)ofthetargetvariableatthegivenvaluesoftheinputvariables.Becausethepredictionsareprobabilities,whichareboundedby0and1andarenotlinearinthi
5、sspace,theprobabilitiesmustbetransformedinordertobeadequatelymodeled.Themostcommontransformationforabinarytargetisthelogittransformation.Probitandcomplementarylog-logtransformationsarealsoavailableintheregressionnode.Error!Notextofspecifiedstyleindocument.Error!Notextofspecifie
6、dstyleindocument.3-53Recallthatoneassumptionoflogisticregressionisthatthelogittransformationoftheprobabilitiesofthetargetvariableresultsinalinearrelationshipwiththeinputvariables.Error!Notextofspecifiedstyleindocument.Error!Notextofspecifiedstyleindocument.3-53Regressionusesonl
7、yfullcasesinthemodel.Thismeansthatanycase,orobservation,thathasamissingvaluewillbeexcludedfromconsiderationwhenbuildingthemodel.Asdiscussedearlier,whentherearemanypotentialinputvariablestobeconsidered,thiscouldresultinanunacceptablyhighlossofdata.Therefore,whenpossible,missingv
8、aluesshouldbeimputedpriortorunningaregressionmodel.Oth
此文档下载收益归作者所有