分层回归分析

《分层回归分析》由会员上传分享，免费在线阅读，更多相关内容在工程资料-天天文库。

1、分层回归分析  2007-12-0814:55:16

2、  分类：专业补充

3、  标签：

4、字号大中小 订阅 HierarchicalRegressionAnalysis Inahierarchicalmultipleregression,theresearcherdecidesnotonlyhowmanypredictorstoenterbutalsotheorderinwhichtheyenter.Usually,theorderofentryisbasedonlogicalortheoreticalconsiderations.Therearethreepredict

5、orvariablesandonecriterionvariableinthefollowingdataset.AresearcherdecidedtheorderofentryisX1,X2,andX3. SPSSforWindows1.EnterData.2.ChooseAnalyze/ Regression/Linear.Dependent:Select"y"andmoveittotheDependentvariablelist.First,clickon thevariabley.Next,clickontherightarrow.Block1of1Indep

6、endent(s):Choosethefirstpredictorvariablex1andmoveittotheIndependent(s)box.Next,clicktheNextbuttonasshownbelow.Block2of2Clickthepredictorvariablex2andmoveittotheIndependent(s)box.Next,clicktheNextbuttonasshownbelow. Block3of3Clickthepredictorvariablex3andmoveittotheIndependent(s)box.3.C

7、licktheStatisticsbutton.CheckRsquaredchange. ClickContinueandOK. SPSSOutput1.RsquareChangeRSquareandRSquareChangeOrderofEntryModel1:EnterX1Model1:Rsquare=.25ThepredictorX1aloneaccountsfor25%ofthevarianceinY.R2=.25Model2:EnterX2 next.Model2:Rsquare=.582TheIncreaseinRsquare:.582-.25=.332T

8、hepredictorX2accountsfor33%ofthevarianceinYaftercontrollingforX1.R2=.25+.332=.582ModelThree:Enter X3 thirdModel3:Rsquare=.835TheIncreaseinRsquare:.835-.582=.253ThepredictorX3accountsfor25%ofthevarianceinY,afterX1andX2werepartialedoutfromX3.R2=.25+.332+.253=.835About84%ofthevarianceinthe

9、criterionvariablewasexplainedbythefirst(25%),second(33%)andthird(25%)predictorvariables. 2.AdjustedRSquareForourexample,thereareonlyfivesubjects.However,therearethreepredictors.RecallthatRsquaremaybeoverestimatedwhenthedatasetshavefewcases(n)relativetonumberofpredictors(k).Datasetswitha

10、smallsamplesizeandalargenumberofpredictorswillhaveagreaterdifferencebetweentheobtainedandadjustedRsquare(.25vs..000,.582vs..165,and.835vs..338). 3.FChangeandSig.FChange IftheRsquarechangeassociatedwithapredictorvariableinquestionislarge,itmeansthatthepredictorvariableisagoodpre