资源描述:
《mcmc methods for multi response generalized linear mixed models the mcmcglmm r package (hadfield, 2010)外语英文电子书》由会员上传分享,免费在线阅读,更多相关内容在教育资源-天天文库。
1、JournalofStatisticalSoftwareJSSJanuary2010,Volume33,Issue2.http://www.jstatsoft.org/MCMCMethodsforMulti-ResponseGeneralizedLinearMixedModels:TheMCMCglmmRPackageJarrodD.HadeldUniversityofEdinburghAbstractGeneralizedlinearmixedmodelsprovidea
exibleframeworkformodelingaran
2、geofdata,althoughwithnon-Gaussianresponsevariablesthelikelihoodcannotbeobtainedinclosedform.MarkovchainMonteCarlomethodssolvethisproblembysamplingfromaseriesofsimplerconditionaldistributionsthatcanbeevaluated.TheRpackageMCMCglmmimplementssuchanalgorithmforarangeofmodelt
3、tingproblems.Morethanoneresponsevariablecanbeanalyzedsimultaneously,andthesevariablesareallowedtofollowGaussian,Poisson,multi(bi)nominal,exponential,zero-in
atedandcensoreddis-tributions.Arangeofvariancestructuresarepermittedfortherandomeects,includinginteractionswithca
4、tegoricalorcontinuousvariables(i.e.,randomregression),andmorecomplicatedvariancestructuresthatarisethroughsharedancestry,eitherthroughapedi-greeorthroughaphylogeny.Missingvaluesarepermittedintheresponsevariable(s)anddatacanbeknownuptosomelevelofmeasurementerrorasinmeta-a
5、nalysis.Allsimu-lationisdoneinC/C++usingtheCSparselibraryforsparselinearsystems.Keywords:MCMC,linearmixedmodel,pedigree,phylogeny,animalmodel,multivariate,sparse,R.Duetotheir
exibility,linearmixedmodelsarenowwidelyusedacrossthesciences(BrownandPrescott1999;PinheiroandBat
6、es2000;Demidenko2004).However,generalizingthesemodelstonon-Gaussiandatahasproveddicultbecauseintegratingovertherandomeectsisintractable(McCullochandSearle2001).Althoughtechniquesthatapproximatetheseintegrals(BreslowandClayton1993)arenowpopular,MarkovchainMonteCarlo(MCM
7、C)methodsprovideanalternativestrategyformarginalizingtherandomeectsthatmaybemorerobust(Zhao,Staudenmayer,Coull,andWand2006;BrowneandDraper2006).DevelopingMCMCmethodsforgeneralizedlinearmixedmodels(GLMM)isanactiveareaofresearch(e.g.,ZegerandKarim1991;Damien,Wakeeld,andW
8、alker1999;SorensenandGianola2002;Zhaoetal.2006),andseveralsoftwarepackagesarenowavailablethatimplementt