PackageloriDecember162020TypePackageTitleImputationofHighDimension - PDF document

1 Package`lori'December16,2020TypePackageT
Package`lori'December16,2020TypePackageTitleImputationofHigh-DimensionalCountDatausingSideInformationVersion2.2.2MaintainerGenevieveRobin&#xgene;&#xviev;.ro;in@;nrs;&#x.fr0;DescriptionAnalysis,imputation,andmultipleimputationofcountdatausingcovariates.LORIusesalog-linearPoissonmodelwheremainrowandcolumneffects,aswellaseffectsofknowncovari-atesandinteractiontermscanbetted.Theestimationprocedureisbasedontheconvexopti-mizationofthePoissonlosspenalizedbyaLassotypepenaltyandanuclearnorm.LORIre-turnsestimatesofmaineffects,covariateeffectsandinteractions,aswellasanimputedcountta-ble.Thepackagealsocontainsamultipleimputationprocedure.Themethodsarede-scribedinRobin,Josse,MoulinesandSardy(2019) rXi;─v:1703.02296v4.LicenseGPL-3EncodingUTF-8LazyDatatrueDependsstats,data.table,rARPACK,svd,R rXi;─(=2.10)Suggestsknitr,rmarkdown,testthatVignetteBuilderknitrRoxygenNote7.1.1NeedsCompilationnoAuthorGenevieveRobin[aut,cre]RepositoryCRANDate/Publication2020-12-1614:00:06UTCRtopicsdocumented:aravo.............................................2covmat.........................

2 ...................3cv.lori.............
...................3cv.lori............................................3lori..............................................5mi.lori............................................71 2aravopool.lori...........................................8qut..............................................9Index11 aravoAlpineplantcommunitiesinAravo,France:Abundancedataandco-variates DescriptionOriginallypublishedinCholer,P.2005.ConsistentshiftsinAlpineplanttraitsalongamesotopo-graphicalgradient.Arctic,Antarctic,andAlpineResearch37:444–453.Usagedata(aravo)FormatAlistwith4attributes:speabundancetableof82speciesin75environmentsenvamatrixof6covariatesforthe75environmentstraitsamatrixof8covariatesforthe82speciesspe.namesavectorof82speciesnamesDetailsAnalysedinDray,S.,Choler,P.,Dolédec,S.,Peres-Neto,P.R.,Thuiler,W.,Pavoine,S.&terBraak,C.J.F.2014.Combiningthefourth-cornerandtheRLQmethodsforassessingtraitresponsestoenvironmentalvariation.Ecology95:14-21DescriptionfromDrayetal.(2014):Communitycompositionofvascularplantswasdeterminedin755×5mplots.Eachsitewasdescribedbysixenvironmentalvariables:meansno

3 wmeltdateovertheperiod1997–1999,slo
wmeltdateovertheperiod1997–1999,slopeinclination,aspect,indexofmicroscalelandform,indexofphysicaldisturbanceduetocryoturbationandsoliuction,andanindexofzoogenicdisturbanceduetotramplingandburrowingactivitiesoftheAlpinemarmot.Allvariablesarequantitativeexceptthelandformandzoogenicdisturbanceindicesthatarecategoricalvariableswithveandthreecategories,respectively.Eightquantitativefunctionaltraits(i.e.,vegetativeheight,lateralspread,leafelevationangle,leafarea,leafthickness,specicleafarea,mass-basedleafnitrogencontent,andseedmass)weremeasuredonthe82mostabundantplantspecies(outofatotalof132recordedspecies).Sourcehttp://pbil.univ-lyon1.fr/ade4/ade4-html/aravo.html covmat3 covmatcovmat DescriptioncovmatUsagecovmat(n,p,R=NULL,C=NULL,E=NULL,center=F)ArgumentsnnumberofrowspnumberofcolumnsRnxK1matrixofrowcovariatesCnxK2matrixofcolumncovariatesE(n+p)xK3matrixofrow-columncovariatescenterbooleanindicatingwhetherthereturnedcovariatematrixshouldbecentered(foridentiability)ValuethejointproductofRandCcolumn-bindedwithE,a(np)x(K1+K2+K3)matrixinorderrow1col1,row2col1,...,rown

4 col1,row1col2,row2col2,...,rowncolpExamp
col1,row1col2,row2col2,...,rowncolpExamplesRmatrix(rnorm(10),5)Cmatrix(rnorm(9),3)covscovmat(5,3,R,C) cv.loriThecv.lorimethodperformsautomaticselectionoftheregularizationparameters(lambda1andlambda2)usedinthelorifunction.Theseparametersareselectedbycross-validation.Theclassicalprocedureistoapplycv.loritothedatatoselecttheregularizationparameters,andtothenimputeandanalyzethedatausingthelorifunction(ormi.loriformultipleimputation). DescriptionThecv.lorimethodperformsautomaticselectionoftheregularizationparameters(lambda1andlambda2)usedinthelorifunction.Theseparametersareselectedbycross-validation.Theclassicalprocedureistoapplycv.loritothedatatoselecttheregularizationparameters,andtothenimputeandanalyzethedatausingthelorifunction(ormi.loriformultipleimputation). 4cv.loriUsagecv.lori(Y,cov=NULL,intercept=T,reff=T,ceff=T,rank.max=5,N=5,len=20,prob=0.2,algo=c("alt","mcgd"),thresh=1e-05,maxit=10,trace.it=F,parallel=F)ArgumentsY[matrix,data.frame]abundancetable(nxp)cov[matrix,data.frame]designmatris(npxq)intercept[boolean]whetheraninterceptshouldbetted,defaultvalueisFALSEreff[bool

5 ean]whetherroweffectsshouldbetted,de
ean]whetherroweffectsshouldbetted,defaultvalueisTRUEceff[boolean]whethercolumneffectsshouldbetted,defaultvalueisTRUErank.max[integer]maximumrankofinteractionmatrix,defaultis2N[integer]numberofcross-validationfoldslen[integer]thesizeofthegridprob[numericin(0,1)]theproportionofentriestoremoveforcross-validationalgotypeofalgorithmtouse,eitheroneof"mcgd"(mixedcoordinategradientde-scent,adaptedtolargedimensions)or"alt"(alternatingminimization,adaptedtosmalldimensions)thresh[positivenumber]convergencethreshold,defaultis1e-5maxit[integer]maximumnumberofiterations,defaultis100trace.it[boolean]whetherinformationaboutconvergenceshouldbeprintedparallel[boolean]whethercomputationsshouldbeperformedinparallelonmultiplecoresValueAlistwiththefollowingelementslambda1regularizationparameterestimatedbycross-validationfornuclearnormpenalty(interactionmatrix) lori5lambda2regularizationparameterestimatedbycross-validationforl1normpenalty(maineffects)errorsatablecontainingthepredictionerrorsforallpairsofparametersExamplesXmatrix(rnorm(20),10)Ymatrix(rpois(10,1:10),5)rescv.lori(Y,X,N=2,len=2)

6 loriThelorimethodimplementsamethodtoana
loriThelorimethodimplementsamethodtoanalyzeandimputeincom-pletecounttables.Animportantfeatureofthemethodisthatitcantakeintoaccountmaineffectsofrowsandcolumns,aswellaseffectsofcontinuousorcategoricalcovariates,andinteraction.Theestima-tionprocedureisbasedonminimizingaPoissonlosspenalizedbyaLassotypepenalty(sparsevectorofcovariateeffects)andanuclearnormpenaltyinducingalow-rankinteractionmatrix(afewlatentfac-torssummarizetheinteractions). DescriptionThelorimethodimplementsamethodtoanalyzeandimputeincompletecounttables.Animportantfeatureofthemethodisthatitcantakeintoaccountmaineffectsofrowsandcolumns,aswellaseffectsofcontinuousorcategoricalcovariates,andinteraction.TheestimationprocedureisbasedonminimizingaPoissonlosspenalizedbyaLassotypepenalty(sparsevectorofcovariateeffects)andanuclearnormpenaltyinducingalow-rankinteractionmatrix(afewlatentfactorssummarizetheinteractions).Usagelori(Y,cov=NULL,lambda1=NULL,lambda2=NULL,intercept=T,reff=T,ceff=T,rank.max=2,algo=c("alt","mcgd"),thresh=1e-05,maxit=100,trace.it=F,parallel=F) 6loriArgumentsY[matrix,data.frame]counttable(nxp).cov[matr

7 ix,data.frame]designmatrix(np*q)inorderr
ix,data.frame]designmatrix(np*q)inorderrow1xcol1,row2xcol2,..,rownxcol1,row1xcol2,row2xcol2,...,...,rownxcolplambda1[positivenumber]theregularizationparameterfortheinteractionmatrix.lambda2[positivenumber]theregularizationparameterforthecovariateeffects.intercept[boolean]whetheraninterceptshouldbetted,defaultvalueisFALSEreff[boolean]whetherroweffectsshouldbetted,defaultvalueisTRUEceff[boolean]whethercolumneffectsshouldbetted,defaultvalueisTRUErank.max[integer]maximumrankofinteractionmatrix(smallerthanmin(n-1,p-1))algotypeofalgorithmtouse,eitheroneof"mcgd"(mixedcoordinategradientde-scent,adaptedtolargedimensions)or"alt"(alternatingminimization,adaptedtosmalldimensions)thresh[positivenumber]convergencetoleranceofalgorithm,bydefault1e-6.maxit[integer]maximumallowednumberofiterations.trace.it[boolean]whetherconvergenceinformationshouldbeprintedparallel[boolean]whethercomputationsshouldbeperformedinparallelonmultiplecoresValueAlistwiththefollowingelementsXnxpmatrixoflogofexpectedcountsalpharoweffectsbetacolumneffectsepsiloncovariateeffectsthetanxpmatrixofrow-columninter

8 actionsimputednxpmatrixofimputedcountsme
actionsimputednxpmatrixofimputedcountsmeansnxpmatrixofexpectedcounts(exp(X))covnpxKmatrixofcovariatesExamples mi.lori7 mi.loriThemi.loriperformsMmultipleimputationsusingthelorimethod.Multipleimputationallowstoproduceestimatesofmissingvalues,aswellasintervalsofvariability.TheclassicalprocedureistoperformMmultipleimputationsusingthemi.lorimethod,andtoaggregatethemusingthepool.lorimethod. DescriptionThemi.loriperformsMmultipleimputationsusingthelorimethod.Multipleimputationallowstoproduceestimatesofmissingvalues,aswellasintervalsofvariability.TheclassicalprocedureistoperformMmultipleimputationsusingthemi.lorimethod,andtoaggregatethemusingthepool.lorimethod.Usagemi.lori(Y,cov=NULL,lambda1=NULL,lambda2=NULL,M=25,intercept=T,reff=T,ceff=T,rank.max=5,algo=c("alt","mcgd"),thresh=1e-05,maxit=1000,trace.it=F)ArgumentsY[matrix,data.frame]counttable(nxp).cov[matrix,data.frame]designmatrix(np*q)inorderrow1xcol1,row2xcol2,..,rownxcol1,row1xcol2,row2xcol2,...,...,rownxcolplambda1[positivenumber]theregularizationparameterfortheinteractionmatrix.lambda2[positivenumber]theregularizationparamete

9 rforthecovariateeffects.M[integer]thenum
rforthecovariateeffects.M[integer]thenumberofmultipleimputationstoperformintercept[boolean]whetheraninterceptshouldbetted,defaultvalueisFALSEreff[boolean]whetherroweffectsshouldbetted,defaultvalueisTRUEceff[boolean]whethercolumneffectsshouldbetted,defaultvalueisTRUErank.max[integer]maximumrankofinteractionmatrix(smallerthanmin(n-1,p-1)) 8pool.lorialgotypeofalgorithmtouse,eitheroneof"mcgd"(mixedcoordinategradientde-scent,adaptedtolargedimensions)or"alt"(alternatingminimization,adaptedtosmalldimensions)thresh[positivenumber]convergencetoleranceofalgorithm,bydefault1e-6.maxit[integer]maximumallowednumberofiterations.trace.it[boolean]whetherconvergenceinformationshouldbeprintedValuemi.imputedalistoflengthMcontainingtheimputedcounttablesmi.alphaa(Mxn)matrixcontaininginrowstheestimatedroweffects(onerowcorre-spondstoonesingleimputation)mi.betaa(Mxp)matrixcontaininginrowstheestimatedcolumneffects(onerowcorre-spondstoonesingleimputation)mi.epsilona(Mxq)matrixcontaininginrowstheestimatedeffectsofcovariates(onerowcorrespondstoonesingleimputation)mi.thetaalistoflengthMcontaini

10 ngtheestimatedinteractionmatricesmi.mual
ngtheestimatedinteractionmatricesmi.mualistoflengthMcontainingtheestimatedPoissonmeansmi.ylistofbootstrappedcounttablesusedfotmultipleimputationYoriginalincompletecounttableExamplesXmatrix(rnorm(50),25)Ymatrix(rpois(25,1:25),5)resmi.lori(Y,X,10,10,2) pool.loriThepool.lorimethodaggregateslorimultipleimputationresults.Mul-tipleimputationallowstoproduceestimatesofmissingvalues,aswellasintervalsofvariability.Theclassicalprocedureistoperformmulti-pleimputationusingthemi.lorimethod,andtoaggregatethemusingthepool.lorimethod. DescriptionThepool.lorimethodaggregateslorimultipleimputationresults.Multipleimputationallowstoproduceestimatesofmissingvalues,aswellasintervalsofvariability.Theclassicalprocedureistoperformmultipleimputationusingthemi.lorimethod,andtoaggregatethemusingthepool.lorimethod.Usagepool.lori(res.mi) qut9Argumentsres.miamultipleimputationresultfromthefunctionmi.loriValuepool.imputealistcontainingthepooledmeans(mean)andvariance(var)oftheimputedvaluespool.alphaalistcontainingthepooledmeans(mean)andvariance(var)oftheroweffectspool.betaalistcontainingthepooledmeans(mean)and

11 variance(var)ofthecolumneffectspool.epsi
variance(var)ofthecolumneffectspool.epsilonalistcontainingthepooledmeans(mean)andvariance(var)ofthecovariateeffectspool.thetaalistcontainingthepooledmeans(mean)andvariance(var)oftheinteractionsExamplesXmatrix(rnorm(50),25)Ymatrix(rpois(25,1:25),5)resmi.lori(Y,X,10,10,2)poolrespool.lori(res) qutautomaticselectionofnuclearnormregularizationparameter DescriptionautomaticselectionofnuclearnormregularizationparameterUsagequt(Y,cov,lambda2=0,q=0.95,N=100,reff=T,ceff=T)ArgumentsYAmatrixofcounts(contingencytable).covA(np)xKmatrixofKcovariatesaboutrowsandcolumnslambda2Apositivenumber,theregularizationparameterforcovariatesmaineffectsqAnumberbetween0and1.Thequantileofthedistributionof$lambda_QUT$totake.NAninteger.Thenumberofparametricbootstrapsamplestodraw.reff[boolean]whetherroweffectsshouldbetted,defaultvalueisTRUEceff[boolean]whethercolumneffectsshouldbetted,defaultvalueisTRUE 10qutValuethevalueof$lambda_QUT$touseinLoRI.ExamplesX=matrix(rnorm(30),15)Y=matrix(rpois(15,1:15),5)lambda=qut(Y,X,10,N=10) Indexdatasetsaravo,2aravo,2covmat,3cv.lori,3lori,5mi.lori,7pool.lori,8qut,