多元方法nag mv fac得分g03ccc1目大物实验报告cl05.pdfVIP

g03–MultivariateMethodsg03ccc

nagmvfacscore(g03ccc)

1.Purpose

nagmvfacscore(g03ccc)computesfactorscorecoefficientsfromtheresultoffittingafactor

analysismodelbyumlikelihoodasperformedbynagmvfactor(g03cac).

2.Specification

#includenag.h

#includenagg03.h

voig_mv_fac_score(Nag_FacScoreMethodmethog_FacRotationrotate,

Integernvar,Integernfac,doublefl[],Integertdfl,doublepsi[],

doublee[],doubler[],Integertdr,doublefs[],Integertdfs,

NagError*fail)

3.Description

Afactoranalysismodelaimstoaccountforthecovariancesamongpvariables,observedonn

individuals,intermsofasmallernumber,k,ofunobservedvariablesorfactors.Thevaluesof

thefactorsforanindividualareknownasfactorscores.nagmvfactor(g03cac)fitsthefactor

analysismodelbyumlikelihoodandreturnstheestimatedfactorloadingmatrix,Λ,andthe

diagonalmatrixofvariancesoftheuniquecomponents,Ψ.Toobtainestimatesofthefactors,ap

ˆ

bykmatrixoffactorscorecoefficients,Φ,isformed.Theestimatedvectoroffactorscores,f,is

thengivenby:

T

ˆ

f=xΦ,

wherexisthevectorofobservedvariablesforanindividual.

Therearetwocommonlyusedmethodsofobtainingfactorscorecoefficients.

Theregressionmethod:

Φ=Ψ−1Λ(I+ΛTΨ−1Λ)−1,

andBartlett’smethod: