stepwise的model history 数据怎么导出
发布网友
发布时间：2022-04-25 16:26
共1个回答
热心网友
时间：2023-10-17 11:27
函数名是REGRESS,逐步回归：stepwiseB=REGRESS(Y,X)returnsthevectorBofregressioncoefficientsinthelinearmodelY=X*B.Xisann-by-pdesignmatrix,withrowscorrespondingtoobservationsandcolumnstopredictorvariables.Yisann-by-1vectorofresponseobservations.[B,BINT]=REGRESS(Y,X)returnsamatrixBINTof95%confidenceintervalsforB.[B,BINT,R]=REGRESS(Y,X)returnsavectorRofresials.[B,BINT,R,RINT]=REGRESS(Y,X)returnsamatrixRINTofintervalsthatcanbeusedtodiagnoseoutliers.IfRINT(i,:)doesnotcontainzero,thenthei-thresialislargerthanwouldbeexpected,atthe5%significancelevel.ThisisevidencethattheI-thobservationisanoutlier.[B,BINT,R,RINT,STATS]=REGRESS(Y,X)returnsavectorSTATScontaining,inthefollowingorder,theR-squarestatistic,theFstatisticandpvalueforthefullmodel,andanestimateoftheerrorvariance.[]=REGRESS(Y,X,ALPHA)usesa100*(1-ALPHA)%confidenceleveltocomputeBINT,anda(100*ALPHA)%significanceleveltocomputeRINT.Xshouldincludeacolumnofonessothatthemodelcontainsaconstantterm.TheFstatisticandpvaluearecomputendertheassumptionthatthemodelcontainsaconstantterm,andtheyarenotcorrectformodelswithoutaconstant.TheR-squarevalueisoneminustheratiooftheerrorsumofsquarestothetotalsumofsquares.Thisvaluecanbenegativeformodelswithoutaconstant,whichindicatesthatthemodelisnotappropriateforthedata.IfcolumnsofXarelinearlydependent,REGRESSsetsthemaximumpossiblenumberofelementsofBtozerotoobtaina"basicsolution",andreturnszerosinelementsofBINTcorrespondingtothezeroelementsofB.REGRESStreatsNaNsinXorYasmissingvalues,andremovesthem.Seealsolscov,polyfit,regstats,robustfit,stepwise.
