September6,2016
Version0.8.1Date2016-09-06
TitleNetworkMeta-AnalysisUsingBayesianMethodsAuthorGertvanValkenhoef,JoelKuiper
MaintainerGertvanValkenhoef DescriptionNetworkmeta-analyses(mixedtreatmentcomparisons)intheBayesian frameworkusingJAGS.Includesmethodstoassessheterogeneityandinconsistency,andanumberofstandardvisualizations.Dependscoda(>=0.13) Importsigraph(>=1.0),meta(>=2.1),plyr(>=1.8),graphics, grDevices,stats,utils,grid,rjags(>=4-0),truncnorm,RglpkSuggeststestthat(>=0.8),Matrix,XML(>=3.6)URLhttp://github.com/gertvv/gemtcLicenseGPL-3LazyDatatrue Collate'anohe.R''arrayize.R''blobbogram.R''template.R''code.R' 'data.R''deviance.R''forest.R''solveLP.R''inits.R''likelihoods.R''ll-helper.counts.R''ll.binom.cloglog.R''ll.binom.log.R''ll.binom.logit.R''ll.call.R''ll.normal.identity.R''ll.poisson.log.R' 'minimum.diameter.spanning.tree.R''mtc.data.studyrow.R''mtc.hy.prior.R''mtc.model.R''mtc.model.consistency.R''mtc.model.nodesplit.R''mtc.model.regression.R''mtc.model.ume.R''mtc.model.use.R''mtc.network.R''mtc.network.xml.R''stopIfNotConsistent.R''mtc.result.R''mtc.run.R''nodesplit.R''plotCovariateEffect.R''priors.R''rank.probability.R''regression.R''relative.effect.R''relative.effect.table.R'RoxygenNote5.0.1NeedsCompilationyesRepositoryCRAN Date/Publication2016-09-0615:34:06 1 2gemtc-package Rtopicsdocumented: gemtc-package..atrialFibrillation..blobbogram....blocker......certolizumab...depression.....dietfat.......hfPrevention....ll.call.......mtc.anohe.....mtc.data.studyrow.mtc.deviance...mtc.hy.prior....mtc.model.....mtc.network....mtc.nodesplit...mtc.run......parkinson.....plotCovariateEffectrank.probability..read.mtc.network.relative.effect...relative.effect.tablesmoking......thrombolytic.... Index .............................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................24581010111214151618222426282930313234353637 gemtc-packageGeMTC:Networkmeta-analysisinR Description AnRpackageforperformingnetworkmeta-analyses(mixedtreatmentcomparisons).Details Networkmeta-analysis,ormixedtreatmentcomparison(MTC)isatechniquetometa-analyzenet-worksoftrialscomparingtwoormoretreatmentsatthesametime[Diasetal.2013a].UsingaBayesianhierarchicalmodel,alldirectandindirectcomparisonsaretakenintoaccounttoarriveatasingleconsistentestimateoftheeffectofallincludedtreatmentsbasedonallincludedstudies.Thispackageallowstheautomatedgenerationofnetworkmeta-analysismodels[vanValkenhoefetal.2012],inclusingbothfixedeffectandrandomeffectsnetworkmeta-analysis,node-splittingmodelstoidentifyinconsistency,andnetworkmeta-regressionmodels.ModelsareestimatedusingJAGS(throughtherjagspackage). gemtc-package3 ThepackagecanreadGeMTCfilesthatwerecreatedbytheGeMTCGUI(nowobsolete)orex-portedfromADDIS1.x,iftheoptionalpackageXMLisinstalled.ThesourceforGeMTCisavailableundertheGPL-3onGithub.Author(s) GertvanValkenhoefReferences S.Dias,N.J.Welton,D.M.Caldwell,andA.E.Ades(2010),Checkingconsistencyinmixedtreat-mentcomparisonmeta-analysis,StatisticsinMedicine29(7-8,Sp.Iss.SI):932-944.[doi:10.1002/sim.3767] S.Dias,A.J.Sutton,A.E.Ades,andN.J.Welton(2013a),AGeneralizedLinearModelingFrame-workforPairwiseandNetworkMeta-analysisofRandomizedControlledTrials,MedicalDecisionMaking33(5):607-617.[doi:10.1177/02729X12458724] S.Dias,A.J.Sutton,N.J.Welton,andA.E.Ades(2013b),Heterogeneity-Subgroups,Meta-Regression,Bias,andBias-Adjustment,MedicalDecisionMaking33(5):618-0.[doi:10.1177/02729X13485157] S.Dias,N.J.Welton,A.J.Sutton,D.M.Caldwell,G.Lu,andA.E.Ades(2013c),InconsistencyinNetworksofEvidenceBasedonRandomizedControlledTrials,MedicalDecisionMaking33(5):1-656.[doi:10.1177/02729X12455847] A.Gelman,A.Jakulin,M.GraziaPittau,Y.-S.Su(2008),Aweaklyinformativedefaultpriordistri-butionforlogisticandotherregressionmodels,TheAnnalsofAppliedStatistics2(4):1360-1383.[doi:10.1214/08-AOAS191] R.M.Turner,J.Davey,M.J.Clarke,S.G.Thompson,J.P.T.Higgins(2012),Predictingtheextentofheterogeneityinmeta-analysis,usingempiricaldatafromtheCochraneDatabaseofSystematicReviews,InternationalJournalofEpidemiology41(3):818-827.[doi:10.1093/ije/dys041]G.vanValkenhoef,G.Lu,B.deBrock,H.Hillege,A.E.Ades,andN.J.Welton(2012),Automatingnetworkmeta-analysis,ResearchSynthesisMethods3(4):285-299.[doi:10.1002/jrsm.10]G.vanValkenhoef,S.Dias,A.E.Ades,andN.J.Welton(2015),Automatedgenerationofnode-splittingmodelsforassessmentofinconsistencyinnetworkmeta-analysis,ResearchSynthesisMethods,acceptedmanuscript.[doi:10.1002/jrsm.1167] G.vanValkenhoefetal.(draft),Modelinginconsistencyasheterogeneityinnetworkmeta-analysis,draftmanuscript. D.E.Warn,S.G.Thompson,andD.J.Spiegelhalter(2002),Bayesianrandomeffectsmeta-analysisoftrialswithbinaryoutcomes:methodsfortheabsoluteriskdifferenceandrelativeriskscales,StatisticsinMedicine21(11):1601-1623.[doi:10.1002/sim.11]SeeAlso mtc.network,mtc.model,mtc.run 4Examples #Loadtheexamplenetworkandgenerateaconsistencymodel:model<-mtc.model(smoking,type=\"consistency\") atrialFibrillation #Loadpre-generatedsamplesinsteadofruningthemodel:##Notrun:results<-mtc.run(model,thin=10) results<-dget(system.file(\"extdata/luades-smoking.samples.gz\package=\"gemtc\"))#Printabasicstatisticalsummaryoftheresults:summary(results) ##Iterations=5010:25000##Thinninginterval=10##Numberofchains=4 ##Samplesizeperchain=2000## ##1.Empiricalmeanandstandarddeviationforeachvariable,##plusstandarderrorofthemean:####MeanSDNaiveSETime-seriesSE##d.A.B0.49650.40810.0045630.0049##d.A.C0.83590.24330.0027200.003147##d.A.D1.10880.43550.0048690.005280##sd.d0.84650.19130.0021390.002965## ##2.Quantilesforeachvariable:####2.5%25%50%75%97.5%##d.A.B-0.29850.23120.49100.75301.341##d.A.C0.38780.67200.82730.98671.353##d.A.D0.26920.81971.09831.38242.006##sd.d0.55090.71190.81800.921.283 atrialFibrillationPreventionofstrokeinatrialfibrillationpatients Description Adatasetof25trialsinvestigating17treatmentsforstrokepreventioninatrialfibrillationpatients.Themainoutcomeisthenumberofpatientswithastroke,andacovariatecapturestheproportionofpatientswithapriorstroke. DataaretakefromTable1ofCooperetal.(2009),withthefollowingcorrectionsapplied:SPAF3andAFASAK2donothaveatreatment13arm,andSPAF1doesnotcontaintreatment5,buttreatment6.Thankstoprof.Cooperforprovidingtheoriginalanalysisdataset.Format Anetworkmeta-regressiondatasetcontaining60rowsofarm-baseddata(respondersandsamplesize). blobbogramSource 5 Cooperetal.(2009),Adressingbetween-studyheterogeneityandinconsistencyinmixedtreatmentcomparisons:Applicationtostrokepreventiontreatmentsinindividualswithnon-rheumaticatrialfibrillation,StatisticsinMedicine28:1861-1881.[doi:10.1002/sim.3594]Examples #BuildamodelsimilartoModel4(b)fromCooperetal.(2009):classes<-list(\"control\"=c(\"01\"), \"anti-coagulant\"=c(\"02\ \"anti-platelet\"=c(\"05\\"mixed\"=c(\"13\regressor<-list(coefficient=shared, variable=stroke,classes=classes)model<-mtc.model(atrialFibrillation, type=\"regression\regressor=regressor,om.scale=10)##Notrun: result<-mtc.run(model)##End(Notrun) blobbogramPlotablobbogram(AKAforestplot) Description blobbogramisaflexiblefunctionforcreatingblobbograms(forestplots),makingnospecificas-sumptionsaboutthedatabeingplotted.Itsupportscolumnandrowgroupingaswellaspagination.Usage blobbogram(data,id.label=\"Study\ci.label=\"Mean(95%CI)\ left.label=NULL,right.label=NULL,center.label=NULL,log.scale=FALSE,xlim=NULL,styles=NULL,grouped=TRUE,group.labels=NULL,columns=NULL,column.labels=NULL, column.groups=NULL,column.group.labels=NULL,digits=2, ask=dev.interactive(orNone=TRUE),draw.no.effect=TRUE)forest(x,...) 6 Arguments dataid.labelci.labelleft.labelright.labelcenter.labellog.scalexlim blobbogram Adataframecontainingonerowforeachconfidenceintervaltobevisualized.Thedataformatisdescribedbelow.Labeltoshowabovetherow-idcolumn.Labeltoshowabovetheconfidenceintervals. Labeltoshowontheleft-handsideoftheno-differenceline.Labeltoshowontheright-handsideoftheno-differenceline.Labeltoshowcenter-alignedwiththeno-differenceline. IfTRUE,theconfidenceintervalsaregivenonalogscale,andaxislabelswillbeexp()transformed. Thescalelimitsoftheplot,iftheconfidenceintervalexceedstheselimitsanarrowwillbeshownatthelimit.Ifunspecified,limitswillbechosenthaten-compassallconfidenceintervals. Adataframedescribingthedifferentrowstyles.Bydefault,thestyles\"normal\\"pooled\"and\"group\"aredefined. IfTRUE,andgroup.labelsarespecified,rowswillbegroupedaccordingtothe\"group\"columngiveninthedataargument.Vectorofgrouplabels. Additionaluser-definedcolumnstobeshown(namesofcolumnsgiveninthedataargument). Avectoroflabelsfortheuser-definedcolumns. stylesgroupedgroup.labelscolumnscolumn.labels column.groupsColumngroups,anumericvectorspecifyingthecolumngroupforeachcolumn.column.group.labels Avectoroflabelsforthecolumngroups.digitsask Thenumberof(significant)digitstoprint. IfTRUE,apromptwillbedisplayedbeforegeneratingthenextpageofamulti-pageplot. Anobjecttocreateaforestplotof.Additionalarguments. draw.no.effectIfTRUE,drawtheno-effectline.x...Details TheforestfunctionisagenericS3method(definitioncompatiblewiththemetapackage).Thispackagedefinesmethodsformtc.resultandmtc.relative.effect.table. Theblobbogramfunctioncreatesablobbogram(forestplot)fromthegivendata(pointestimatesandconfidenceintervals)andmeta-data(labels,columnspecifications,columngroups,rowgroups,styles)usingthegridpackage.Iftheplotwouldnotfitthedevice’sgraphicsregion,thecontentisbrokenupintomultipleplotsgeneratedinsequence(pagination). Thedataargumentisgivenasadataframecontainingthefollowingcolumns:•id:identifier(label)forthisrow. blobbogram •group(optional):rowgroupthisrowbelongsto(indexesintothegroup.labelsargument).•pe:pointestimate. •ci.l:lowerconfidenceintervallimit.•ci.u:upperconfidenceintervallimit. •style:thestyletoapplytothisrow(definedinthestylesargument). 7 Additionaluser-definedcolumnscanbespecifiedusingthecolumnsandcolumn.labelsargu-ments. Thestylesargumentisgivenasadataframecontainingthefollowingcolumns:•style:nameofthestyle.•weight:fontweight. •pe.style:symboltodrawforthepointestimate(\"circle\"or\"square\currently). Value None.Note Thismethodshouldnotbeconsideredstable.Weintendtogeneralizeitfurtherandpossiblyprovideitinaseparatepackage.Theinterfacemaychangeatanytime.Author(s) GertvanValkenhoef,JoëlKuiperSeeAlso meta::forest,grid::GridExamples data<-read.table(textConnection(idgrouppeci.lci.ustylevalue.Avalue.B\"Study1\"10.350.080.92\"normal\"\"2/46\"\"7/46\"\"Study2\"10.430.151.14\"normal\"\"4/50\"\"8/49\"\"Study3\"20.310.070.74\"normal\"\"2/97\"\"10/100\"\"Study4\"20.860.342.90\"normal\"\"9/104\"\"6/105\"\"Study5\"20.330.100.72\"normal\"\"4/74\"\"14/74\"\"Study6\"20.470.230.91\"normal\"\"11/120\"\"22/129\"\"Pooled\"NA0.420.151.04\"pooled\"NANA),header=TRUE) data$pe<-log(data$pe) data$ci.l<-log(data$ci.l)data$ci.u<-log(data$ci.u) blobbogram(data,group.labels=c(GROUP1,GROUP2), columns=c(value.A,value.B),column.labels=c(r/n,r/n), 8 column.groups=c(1,2),grouped=TRUE, column.group.labels=c(Intervention,Control), id.label=\"Trial\ci.label=\"OddsRatio(95%CrI)\log.scale=TRUE) certolizumab blockerBetablockerstopreventmortalityaftermyocardialinfarction Description Adatasetof22trialsinvestigatingbetablockersversuscontroltopreventmortalityaftermyocardialinfarction.Numberofeventsandsamplesize.Format Ameta-analysisdatasetcontaining44rowsofarm-baseddata(respondersandsamplesize).Source S.Dias,A.J.Sutton,A.E.Ades,andN.J.Welton(2013a),AGeneralizedLinearModelingFrame-workforPairwiseandNetworkMeta-analysisofRandomizedControlledTrials,MedicalDecisionMaking33(5):607-617.[doi:10.1177/02729X12458724] J.B.Carlin(1992),Meta-analysisfor2×2tables:aBayesianapproach,StatisticsinMedicine11(2):141-158.[doi:10.1002/sim.4780110202] certolizumabCertolizumabPegol(CZP)forRheumatoidArthritis Description Adatasetof12trialsinvestigating6treatmentsandplaceboforrheumatoidarthritis.Themainoutcomeisthenumberofpatientswhoimprovedbyatleast50%ontheAmericanCollegeofRheumatologyscale(ACR50)at6Months.Acovariateispresentforthemeandiseasedurationatbaseline(years).Format Anetworkmeta-regressiondatasetcontaining24rowsofarm-baseddata(respondersandsamplesize).Source S.Dias,A.J.Sutton,N.J.Welton,andA.E.Ades(2013b),Heterogeneity-Subgroups,Meta-Regression,Bias,andBias-Adjustment,MedicalDecisionMaking33(5):618-0.[doi:10.1177/02729X13485157] depressionExamples #RunREregressionmodelwithinformativeheterogeneitypriorregressor<-list(coefficient=shared, variable=diseaseDuration,control=Placebo) #sd~half-Normal(mean=0,sd=0.32) hy.prior<-mtc.hy.prior(type=\"std.dev\distr=\"dhnorm\0,9.77)model<-mtc.model(certolizumab, type=\"regression\regressor=regressor,hy.prior=hy.prior)##Notrun: result<-mtc.run(model)##End(Notrun) 9 depressionTreatmentresponseinmajordepression Description Adatasetof111trialsinvestigating12treatmentsformajordepressionontreatmentresponse.Treatmentresponsewasdefinedasareductionofatleast50%fromthebaselinescoreontheHAM-DorMADRSatweek8(or,ifnotavailable,anothertimebetweenweek6and12). Format Anetworkmeta-analysisdatasetcontaining224rowsofarm-baseddata(respondersandsamplesize). Source Ciprianietal.(2009),Comparativeefficacyandacceptabilityof12new-generationantidepres-sants:amultiple-treatmentsmeta-analysis,Lancet373(9665):746-758.[doi:10.1016/S0140-6736(09)60046-5] 10hfPrevention dietfatEffectsoflow-fatdietsonmortality Description Adatasetof10trialsinvestigatinglow-fatdietversuscontroldietformortality.Numberofeventsandexposureinperson-years.Format Ameta-analysisdatasetcontaining20rowsofarm-baseddata(responders,exposure,andsamplesize).Source S.Dias,A.J.Sutton,A.E.Ades,andN.J.Welton(2013a),AGeneralizedLinearModelingFrame-workforPairwiseandNetworkMeta-analysisofRandomizedControlledTrials,MedicalDecisionMaking33(5):607-617.[doi:10.1177/02729X12458724] Hooperetal.(2000),Reducedormodifieddietaryfatforpreventingcardiovasculardisease,CochraneDatabaseofSystematicReviews2:CD002137.[doi:10.1002/14651858.CD002137] hfPrevention Statinsversusplaceboinprimaryandsecondarypreventionofheartfailure Description Adatasetof19trialscomparingstatinsversusplaceboorusualcareforcholesterollowering.Themainoutcomeisthenumberofdeaths.Trialsareeitherprimaryprevention(nopreviousheartdisease;secondary=0)orsecondaryprevention(previousheartdisease;secondary=1).Format Ameta-regressiondatasetcontaining38rowsofarm-baseddata(respondersandsamplesize).Source S.Dias,A.J.Sutton,N.J.Welton,andA.E.Ades(2013b),Heterogeneity-Subgroups,Meta-Regression,Bias,andBias-Adjustment,MedicalDecisionMaking33(5):618-0.[doi:10.1177/02729X13485157] ll.callExamples #BuildamodelsimilartoProgram1(a)fromDiasetal.(2013b):regressor<-list(coefficient=shared, variable=secondary,control=control)model<-mtc.model(hfPrevention, type=\"regression\regressor=regressor, hy.prior=mtc.hy.prior(\"std.dev\\"dunif\0,5))##Notrun: result<-mtc.run(model)##End(Notrun) 11 ll.callCallalikelihood/link-specificfunction Description GeMTCimplementsvariouslikelihood/linkcombinations.Functionalityspecifictothelikeli-hood/linkishandledbymethodswithnamesendingin. ll.call(fnName,model,...)Arguments fnNamemodel...Details Thefollowingmethodscurrentlyneedtobeimplementedtoimplementalikelihood/link:•mtc.arm.mle:calculatesa(corrected)maximumlikelihoodestimateforanarm-leveleffect.Usedtogeneratestartingvalues. •mtc.rel.mle:calculatesa(corrected)maximumlikelihoodestimateforarelativeeffect.Usedtogeneratestartingvalues. •mtc.code.likelihood:generatesJAGScodeimplementingthelikelihood.•scale.log:returnsTRUEifplotsshouldusethelogscale.•scale.name:returnstheuser-facingnameoftheoutcomemetric. Thenameofthefunctiontocall.Seedetailsforavailablefunctions. AnobjectofS3classmtc.modeldescribinganetworkmeta-analysismodel,oralistcontainingelementsnamed’likelihood’and’link’.Additionalargumentstobepassedtothefunction. 12mtc.anohe •scale.limit.inits:returnsanupperandlowerboundfortheinitialvalues,becausesomeinitialvaluesmighttriggerboundaryconditionssuchasprobability0or1forthebinomial.•required.columns.ab:returnstherequiredcolumnsforarm-baseddata. Thefirsttwomethodscannowalsobeusedtoselectivelyapplycontinuitycorrectionsincasethemaximumlikelihoodestimatesareusedforotherpurposes.mtc.arm.mlehasanadditionalk=0.5argumenttospecifythecorrectionfactor.mtc.rel.mlehasargumentscorrection.force=TRUEtoforceapplicationofthecontinuitycorrectionevenifunnecessary,correction.type=\"constant\"tospecifythetypeofcorrection(specify\"reciprocal\")foracorrectionproportionaltotherecip-rocalofthesizeoftheotherarm,andcorrection.magnitude=1tospecifythe(total)magnitudeofthecorrection.Thesecorrectionsapplyonlyforcountdata,andwillbeignoredforcontinuouslikelihood/links. Value Thereturnvalueofthecalledfunction.Author(s) GertvanValkenhoefSeeAlso mtc.modelExamples #The\"model\"maybeastub. model<-list(likelihood=\"poisson\link=\"log\")ll.call(\"scale.name\model)#\"HazardRatio\" ll.call(\"mtc.arm.mle\model,c(responders=12,exposure=80))#meansd#-1.85629800.1118034 mtc.anoheAnalysisofheterogeneity(ANOHE) Description (EXPERIMENTAL)Generateananalysisofheterogeneityforthegivennetwork.Threetypesofmodelareestimated:unrelatedstudyeffects,unrelatedmeaneffects,andconsistency.Outputofthesummaryfunctioncanpassedtoplotforavisualrepresentation.Usage mtc.anohe(network,...) mtc.anoheArguments network... AnobjectofS3classmtc.network. 13 Argumentstobepassedtomtc.runormtc.model.Thiscanbeusedtosetthelikelihood/linkorthenumberofiterations,forexample. Details Analysisofheterogeneityisintendedtobeaunifiedsetofstatisticsandavisualdisplaythatallowsthesimultaneousassessmentofbothheterogeneityandinconsistencyinnetworkmeta-analysis[vanValkenhoefetal.2014b(draft)]. mtc.anohereturnstheMCMCresultsforallthreetypesofmodel.Togetappropriatesummarystatistics,callsummary()ontheresultsobject.Thesummarycanbeplotted. TocontrolparametersoftheMCMCestimation,seemtc.run.Tospecifythelikelihood/linkortocontrolothermodelparameters,seemtc.model.The...argumentsarefirstmatchedagainstmtc.run,andthosethatdonotmatcharepassedtomtc.model.Value Formtc.anohe:anobjectofclassmtc.anohe.Thisisalistwiththefollowingelements:result.useresult.umeresult.cons TheresultfortheUSEmodel(seemtc.run).TheresultfortheUMEmodel(seemtc.run).Theresultfortheconsistencymodel(seemtc.run). Forsummary:anobjectofclassmtc.anohe.summary.Thisisalistwiththefollowingelements:cons.modelstudyEffectspairEffectsconsEffectsindEffectsisquared.compisquared.globNote Thismethodshouldnotbeconsideredstable.Itisanexperimentalfeatureandheavilyworkinprogress.Theinterfacemaychangeatanytime.Author(s) GertvanValkenhoef,JoëlKuiperSeeAlso mtc.modelmtc.run Generatedconsistencymodel. Study-leveleffectsummaries(multi-armtrialsdownweighted).Pair-wisepooledeffectsummaries(fromtheUMEmodel).Consistencyeffectsummaries. Indirecteffectsummaries(back-calculated).Per-comparisonI-squaredstatistics.GlobalI-squaredstatistics. 14mtc.data.studyrow mtc.data.studyrowConvertone-study-per-rowdatasets Description Convertsdatasetsintheone-study-per-rowformattoone-arm-per-rowformatusedbyGeMTCUsage mtc.data.studyrow(data, armVars=c(treatment=t,responders=r,sampleSize=n),nArmsVar=na,studyVars=c(), studyNames=1:nrow(data),treatmentNames=NA, patterns=c(%s..,%s..%d.))Arguments dataarmVars Datainone-study-per-rowformat. Vectorofper-armvariables.Thenameofeachcomponentwillbethecolumnnameintheresultingdataset.Thecolumnnameinthesourcedatasetisderivedfromthevalueofeachcomponent. Variableholdingthenumberofarmsforeachstudy. Vectorofper-studyvariables.Thenameofeachcomponentwillbethecolumnnameintheresultingdataset.Thecolumnnameinthesourcedatasetisderivedfromthevalueofeachcomponent.Vectorofstudynames. Patternstogeneratecolumnnamesinthesourcedataset.Thefirstisforper-studyvariables,thesecondforper-armvariables. nArmsVarstudyVars studyNamespatterns treatmentNamesVectoroftreatmentnames. Details Mapstheone-study-per-rowformatthatiswidelyusedandconvenientforBUGSmodelstotheone-arm-per-rowformatusedbyGeMTC.AstheprimarypurposeistoinputdatasetsfromBUGSmodels,thedefaultsworkforthestandardBUGSdatatableformat.Inmostcases,itshouldbepossibletojustcopy/pastetheBUGSdatatable(withoutthefinal’END’)andread.tableitintoR,thenapplymtc.data.studyrow.Inmanycases,theresultingtablecanbeprocesseddirectlybymtc.network.Value Adatatablewiththerequestedcolumns. mtc.devianceAuthor(s) GertvanValkenhoefSeeAlso mtc.networkExamples ##ExampletakenfromtheNICEDSUTSDseriesinEvidenceSynthesis,#2##DopamineagonistsforthetreatmentofParkinsons #Readthebugs-formatteddata data.src<-read.table(textConnection( t[,1]t[,2]t[,3]y[,1]y[,2]y[,3]se[,1]se[,2]se[,3]na[]13NA-1.22-1.53NA0.5040.439NA212NA-0.7-2.4NA0.2820.258NA2 124-0.3-2.6-1.20.5050.5100.478334NA-0.24-0.59NA0.2650.3NA234NA-0.73-0.18NA0.3350.442NA245NA-2.2-2.5NA0.1970.190NA2 45NA-1.8-2.1NA0.2000.250NA2),header=TRUE)#Convertthedata,settingtreatmentnamesdata<-mtc.data.studyrow(data.src, armVars=c(treatment=t,mean=y,std.err=se),treatmentNames=c(Placebo,DA1,DA2,DA3,DA4))#Checkthatthedataarecorrectprint(data) #Createanetwork network<-mtc.network(data) 15 mtc.devianceInspectresidualdeviance Description InspecttheposteriorresidualdevianceandsummarizeitusingplotsUsage mtc.deviance(result)mtc.devplot(x,...)mtc.levplot(x,...) ##S3methodforclassmtc.devianceplot(x,auto.layout=TRUE,...) 16Arguments resultx auto.layout...Details Anobjectofclassmtc.result.Anobjectofclassmtc.deviance. mtc.hy.prior IfTRUE,theseparateplotswillbeshownaspanelsonasinglepage.Graphicalparameters. mtc.devplotwillgenerateastemplotoftheposteriordevianceperarm(ifthereareonlyarm-baseddata)orthemeanperdatapointdevianceperstudy(iftherearecontrast-baseddata). mtc.levplotwillplottheleverageversusthesquarerootoftheresidualdeviance(meanperdatapointforeachstudy). Thegenericplotfunctionwilldisplaybothonasinglepage(unlessauto.layout=FALSE).Value mtc.deviancereturnsthedeviancestatisticsofamtc.result.Author(s) GertvanValkenhoefSeeAlso mtc.run mtc.hy.priorSetpriorsfortheheterogeneityparameter Description Thesefunctionsgeneratepriorsfortheheterogeneityparameterinmtc.model.Priorscanbesetexplicitlyor,foroutcomesonthelogodds-ratioscale,basedonempiricalresearch.Usage mtc.hy.prior(type,distr,...) mtc.hy.empirical.lor(outcome.type,comparison.type) mtc.hy.priorArguments typedistr 17 Typeofheterogeneityprior:’std.dev’,’var’,or’prec’forstandarddeviation,variance,orprecisionrespectively. Priordistributionname(JAGSsyntax).Typicaloneswouldbe’dunif’(uni-form),’dgamma’(Gamma),or’dlnorm’(log-normal).Use’dhnorm’forthehalf-normal.Notethat,asinJAGS,theprecision(andnotthevarianceorstan-darddeviation)isusedforthenormaldistributionanditsderivatives. Argumentstothedistr.Canbenumericalvaluesor\"om.scale\"fortheesti-matedoutcomemeasurescale(seemtc.model) Thetypeofoutcometogetanempiricalpriorfor.Canbeoneof’mortality’(all-causemortality),’semi-objective’(e.g.cause-specificmortality,majormor-bidityevent,drop-outs),or’subjective’(e.g.pain,mentalhealth,dichotomousbiomarkers). Thetypeofcomparisontogetanempiricalpriorfor.Canbeoneof’pharma-control’(pharmacologicalinterventionsversuscontrol),’pharma-pharma’(phar-macologicalversuspharmacologicalinterventions)and’non-pharma’(anyothercomparisons). ... outcome.type comparison.type Details Thegeneratedpriorisalist,thestructureofwhichmaychangewithoutnotice.ItcanbeconvertedtoJAGScompatiblecodeusingas.character. Empiricalpriorsforthelogodds-ratio(LOR)aretakenfrom[Turneretal.2012].Value Avaluetobepassedtomtc.model.Author(s) GertvanValkenhoefSeeAlso mtc.modelExamples #NOTE:themtc.runcommandsbelowareforillustrativepurposes,suchasmall#numberofiterationsshouldobviouslynotbeusedinpractice. #setauniformpriorstandarddeviation model1<-mtc.model(smoking,hy.prior=mtc.hy.prior(\"std.dev\\"dunif\0,2))result<-mtc.run(model1,n.adapt=10,n.iter=10) #setanempirical(log-normal)prioronthevariance model2<-mtc.model(smoking,hy.prior=mtc.hy.empirical.lor(\"subjective\\"non-pharma\")) 18 result<-mtc.run(model2,n.adapt=10,n.iter=10) mtc.model #setagammapriorontheprecision model3<-mtc.model(smoking,hy.prior=mtc.hy.prior(\"prec\\"dgamma\0.01,0.01))result<-mtc.run(model3,n.adapt=10,n.iter=10) mtc.modelGeneratenetworkmeta-analysismodels Description Themtc.modelfunctiongeneratesnetworkmeta-analysismodelsfromanmtc.networkobject.Usage mtc.model(network,type=\"consistency\factor=2.5,n.chain=4,likelihood=NULL,link=NULL,linearModel=\"random\ om.scale=NULL,hy.prior=mtc.hy.prior(\"std.dev\\"dunif\0,\"om.scale\"),dic=TRUE,powerAdjust=NA,...)Arguments networktypefactorn.chainlikelihoodlinklinearModelom.scale AnobjectofS3classmtc.network Astringliteralindicatingthetypeofmodel(allowedvaluesare\"consistency\\"regression\\"nodesplit\\"ume\or\"use\").VariancescalingfactorforthestartingvaluesNumberofchainsinthemodel Thelikelihoodtobeused.Ifunspecified,asuitablelikelihoodwillbeinferredforthegivendata. Thelinkfunctiontobeused.Ifunspecified,asuitablelinkfunctionwillbeinferredforthegivendata. Thetypeoflinearmodeltobegenerated.Canbe\"random\"forarandomeffectsmodel,or\"fixed\"forafixedeffectmodel. Outcomemeasurescale.Representsa\"verylarge\"differenceontheanalysis’outcomescale.Thisisusedtosetvaguepriors.Forthelogodds-ratio,valuesbetween2and5areconsideredreasonable.Forcontinuousoutcomes,thisde-pendsheavilyonthespecificoutcome.Ifleftunspecified,itisdeterminedfromthedata. Heterogeneityprior.Seemtc.hy.prior. WhensettoTRUE,devianceandfittedvalueswillbemonitoredtoallowcom-putationoftheDevianceInformationCriterion(DIC)atresidual. Optional:thenameofacolumninthestudiesdataframeofthemtc.network.Thiscolumnmustcontainvaluesbetween0and1.Thelikelihoodforeachstudywillbeadjustedbyinflatingthevariance,where0meansthestudyisexcludedand1meansitreceivesfullweight.Seedetailsformore. Additionalargumentstobepassedtothetype-specificmodelgenerationfunc-tion. hy.priordic powerAdjust ... mtc.modelDetails 19 Themtc.modelfunctiongeneratesanobjectofS3classmtc.model,whichcanbevisualizedbythegenericplotfunctionorsummarizedbythegenericsummaryfunction.Theselikelihood/linksaresupported: •normal/identity:forcontinuous(meandifference)data. Requiredcolumns:[mean,std.err]or[mean,std.dev,sampleSize].Result:relativemeandifference. •binom/logit:fordichotomousdata. Requiredcolumns[responders,sampleSize].Result:(log)oddsratio. •binom/log:fordichotomousdata. Requiredcolumns[responders,sampleSize].Result:(log)riskratio. •binom/cloglog:forrate(survival)data-equalfollow-upineacharm.Requiredcolumns[responders,sampleSize].Result:(log)hazardratio. •poisson/log:forrate(survival)data. Requiredcolumns[responders,exposure].Result:(log)hazardratio. Mostlikelihood/linksfollow[Diasetal.2013a],andthebinom/logmodelfollows[Warnetal.2002]. Thefollowingmodeltypesaresupported: •consistency:ordinaryconsistencymodel.Noadditionalparameters.[Diasetal.2013a,vanValkenhoefetal.2012] •nodesplit:node-splittingmodel.Removesbotharmsusedtoestimatethedirectevidencefromthenetworkofindirectevidence,ratherthanjustoneofthosearms.Thismeansthatthree-armtrialsdonotcontributeanyevidenceinthenetworkofindirectevidence.Whenrelativeeffectdataarepresent,thesearetransformedappropriately(usinganassumptionofnormality)toenablethisdirect/indirectevidencesplit.Additionalparameters:t1andt2,whichindicatethecomparisontobesplit.[Diasetal.2010,vanValkenhoefetal.2015]•regression:meta-regressionmodel.Additionalparameters:regressor,whichindicateshowtostructurethetreatment-interactionmodelfortheregression.Seebelowfordetails.•use:unrelatedstudyeffects.Modelstheeffectswithineachstudyasifthestudiesareinde-pendent.Noadditionalparameters.[vanValkenhoefetal.(draft)] •ume:unrelatedmeaneffects.Modelstheeffectswithineachcomparisonasiftheyareinde-pendent.Doesnotproperlyhandlemulti-armtrials,andwarnswhentheyarepresentinthenetwork.Noadditionalparameters.[Diasetal.2013b,vanValkenhoefetal.(draft)]Regressorspecification:alistwithelements:variable,coefficient,andeithercontrolorclasses.Thevariableisthenameofthecovariatetoincludeintheregressionanalysis,andmustbeacolumnofthestudiesdataframeinthenetwork.Theregressorvariableisautomati-callycenteredandstandardizedusingthemethodrecommendedbyGelmanetal.(2008).The 20mtc.model coefficientindicatesthetypeoftreatment-interactionmodel:\"shared\\"unrelated\or\"ex-changeable\".control,ifspecified,mustbetheIDofatreatmentinthenetwork.Allothertreat-mentshaveacoefficientrelativetothecontrol,whichcanbethesameforalltreatments(\"shared\"),differentforalltreatments(\"unrelated\")orexchangeablebetweentreatments(\"exchangeable\").classesisanamedlistoftreatmentclasses,thefirstofwhichwillactasthecontrolclass.EachclassisavectoroftreatmentIDsfromthenetwork.Only\"shared\"coefficientscancurrentlybeused,meaningasinglecoefficientperclass(exceptthecontrolclass).SeealsoatrialFibrillation,certolizumab,andhfPreventionforexamplesofmeta-regressionanalyses. StudiescanbedownweightedbyusingthepowerAdjustargument,whichappliesavarianceinfla-tion(alsoknownas\"powerprior\")tothelikelihood.Thisallowsaweightαi∈[0,1]tobespecifiedforeachstudyi.Thelog-likelihoodwillbemultipliedbyafactorαi,orequivalentlyfornormaldistributionsthevariancewillbemultipliedby1/αi.Settingαi=0willcompletelyexcludethatstudy,whereassettingαi=1willweightitfully.Essentially,down-weightedmodelsmodifythedataandhencemodelfitstatisticssuchasDICcannotbecomparedbetweenmodelswithdifferentweightings. Value Anobjectofclassmtc.model.Thefollowingelementsaredescriptive:typenetworktreevar.scalelikelihoodlinkom.scaleregressor Thetypeofmodel NetworkthemodelwasgeneratedfromSpanningtreeformedbythebasicparametersThescalingfactorusedtoover-dispersestartingvaluesThelikelihoodusedThelinkfunctionused Thescaleforthevarianceparameters Regressorspecification(regressionmodelsonly):includesadditionalelements\"center\"and\"scale\"describinghowtheregressorinputwasstandardized TheseelementsdeterminethemodelrunbyJAGS:n.chaincodedatainitsmonitorsdic Thenumberofchains ModelcodeinJAGSsyntax.Usecat()forproperformatting.DatainJAGScompatibleformatInitialvaluesinJAGScompatibleformatThenodesoftheJAGSmodeltomonitorWhethertocomputetheDIC Theselatterfieldscanbemodifiedtoalterthestatisticalmodel,butsuchchangesmaybreakthemodelorassumptionsmadeelsewhereinthepackage. mtc.modelVisualization 21 CallingthegenericplotmethodonanS3mtc.modelobjectwillshowagraphwiththetreatmentsasverticesandthecomparisonsasedges.Thelineswithsolidarrowsrepresentbasicparameters,andtheotherlinesrepresentcomparisonsthatarenotassociatedwithanyparameterbutdohavedirectevidencefromtrials. Theexamplecodewillgeneratethefollowinggraph: BCADEThedefaultlayoutalgorithmisigraph::layout.circle,otherlayoutalgorithmscanbeusedby passingthemasanoptionalargumentcalledlayouttoplot.Theigraph::layout.fruchterman.reingoldalgorithmalsoseemstoproduceniceresultsandmaybebetterforlargegraphs.Author(s) GertvanValkenhoef,JoëlKuiperSeeAlso mtc.network,mtc.runExamples #RandomeffectsconsistencymodelforParkinsonnetworkmodel<-mtc.model(parkinson)plot(model)summary(model) #Fixedeffectmeta-regressionforheartfailurepreventionregressor<-list(coefficient=shared, variable=secondary,control=control) 22 model<-mtc.model(hfPrevention, type=\"regression\regressor=regressor,linearModel=\"fixed\") mtc.network mtc.networkCreateanmtc.network Description Createsanobjectofclassmtc.networkUsage mtc.network(data.ab=data,treatments=NULL,description=\"Network\ data.re=NULL,studies=NULL,data=NULL) ##S3methodforclassmtc.network plot(x,layout=igraph::layout.circle,dynamic.edge.width=TRUE,use.description=FALSE,...)Arguments data.ab Arm-leveldata.Adataframedefiningthearmsofeachstudy,containingthecolumns‘study’and‘treatment’,where‘treatment’mustrefertoanexistingtreatmentIDiftreatmentswerespecified.Furthercolumnsdefinethedataperarm,anddependonthelikelihood/linktobeused.Seemtc.modelforsupportedlikelihood/linksandtheirdatarequirements. Relativeeffectdata.Adataframedefiningthearmsofeachstudy,containingthecolumns‘study’and‘treatment’,where‘treatment’mustrefertoanexistingtreatmentIDiftreatmentswerespecified.Thecolumn‘diff’specifiesthemeandifferencebetweenthecurrentarmandthebaselinearm;set‘diff=NA’forthebaselinearm.Thecolumn‘std.err’specifiesthestandarderrorofthemeandifference(fornon-baselinearms).Fortrialswithmorethantwoarms,specifythestandarderrorofthemeanofthebaselinearmin‘std.err’,asthisdeterminesthecovarianceofthedifferences. Optional.Adataframewithcolumns‘id’and‘description’definingthetreat-mentsoravectorgivingthetreatmentIDs. Optional.Adataframewitha‘study’columnnamingthestudiesandadditionalcolumnscontainingcovariatevalues.Optional.Shortdescriptionofthenetwork. Deprecated.Arm-leveldata;automaticallyassignedtodata.abifitisnotspeci-fied.Presentforcompatibilitywitholderversions.Anmtc.networkobject. data.re treatmentsstudiesdescriptiondatax layoutAnigraph-compatiblelayout.dynamic.edge.width IfsettoTRUE,dynamicallysettheedgewidthbasedonthenumberofstudies. mtc.network use.description DisplaytreatmentdescriptionsinsteadoftreatmentIDs. ...Details One-armtrialsareautomaticallyremoved,whichresultsinawarning. Additionalargumentspassedtoplot.igraph. 23 Alsoseemtc.data.studyrowforaconvenientwaytoimportdatafromtheone-study-per-rowformat,whichisverypopularforBUGScode.Value Formtc.network,anobjectoftheclassmtc.networkwhichisalistcontaining:descriptiontreatmentsdata.abdata.restudies AshortdescriptionofthenetworkAdataframedescribingthetreatments Adataframecontainingthenetworkdata(arm-level)Adataframecontainingthenetworkdata(relativeeffects)Adataframecontainingstudy-levelinformation(covariates) Thesearecleanedupandstandardizedversionsoftheargumentsprovided,orgenerateddefaultsfor‘treatments’iftheargumentwasomitted.Author(s) GertvanValkenhoef,JoëlKuiperSeeAlso mtc.data.studyrow,mtc.modelExamples #Createanewnetworkbyspecifyingallinformation.treatments<-read.table(textConnection(iddescriptionA\"TreatmentA\"B\"TreatmentB\" C\"TreatmentC\"),header=TRUE)data<-read.table(textConnection( studytreatmentresponderssampleSize01A210001B510002B611002C111003A36003C48003B780),header=TRUE) network<-mtc.network(data,description=\"Examplereatments=treatments)plot(network) 24mtc.nodesplit #Createanewnetworkbyspecifyingonlythedata.data<-read.table(textConnection( studytreatmentmeanstd.devsampleSize01A-1.120.61501B-1.550.51602A-0.80.73302B-1.10.531),header=TRUE)network<-mtc.network(data) #Printthenetworkprint(network) ##MTCdataset:Network ##studytreatmentmeanstd.devsampleSize##11A-1.120.615##21B-1.550.516##32A-0.800.733##42B-1.100.531 mtc.nodesplitNode-splittinganalysisofinconsistency Description Generateandrunanensembleofnode-splittingmodels,resultsofwhichcanbejointlysummarizedandplotted.Usage mtc.nodesplit(network,comparisons=mtc.nodesplit.comparisons(network),...)mtc.nodesplit.comparisons(network)Arguments networkcomparisons... AnobjectofS3classmtc.network. Dataframespecifyingthecomparisonstobesplit.Theframehastwocolumns:’t1’and’t2’. Argumentstobepassedtomtc.runormtc.model.Thiscanbeusedtosetthelikelihood/linkorthenumberofiterations,forexample. Details mtc.nodesplitreturnstheMCMCresultsforallrelevantnode-splittingmodels[vanValkenhoefetal.2015].Togetappropriatesummarystatistics,callsummary()ontheresultsobject.Thesummarycanbeplotted.Seemtc.modelfordetailsonhowthenode-splittingmodelsaregenerated.TocontrolparametersoftheMCMCestimation,seemtc.run.Tospecifythelikelihood/linkortocontrolothermodelparameters,seemtc.model.The...argumentsarefirstmatchedagainstmtc.run,andthosethatdonotmatcharepassedtomtc.model. mtc.nodesplit25 mtc.nodesplit.comparisonsreturnsadataframeenumeratingallcomparisonsthatcanreason-ablybesplit(i.e.haveindependentindirectevidence).Value Formtc.nodesplit:anobjectofclassmtc.nodesplit.Thisisalistwiththefollowingelements:d.X.Yconsistency Foreachcomparison(t1=X,t2=Y),theMCMCresultsTheconsistencymodelresults Forsummary:anobjectofclassmtc.nodesplit.summary.Thisisalistwiththefollowingele-ments:dir.effectind.effectcons.effectp.valuecons.modelAuthor(s) GertvanValkenhoef,JoëlKuiperSeeAlso mtc.modelmtc.runExamples #Runallrelevantnode-splittingmodels ##Notrun:result.ns<-mtc.nodesplit(parkinson,thin=50)#(readresultsfromfileinsteadofrunning:) result.ns<-readRDS(system.file(extdata/parkinson.ns.rds,package=gemtc))#Listtheindividualmodelsnames(result.ns) #Timeseriesplotsandconvergencediagnosticsford.A.Cmodelplot(result.ns$d.A.C) gelman.diag(result.ns$d.A.C,multivariate=FALSE)#Overallsummaryandplot summary.ns<-summary(result.ns)print(summary.ns)plot(summary.ns) SummaryofdirecteffectsforeachsplitcomparisonSummaryofindirecteffectsforeachsplitcomparison SummaryofconsistencymodeleffectsforeachsplitcomparisonInconsistencyp-valuesforeachsplitcomparisonThegeneratedconsistencymodel 26mtc.run mtc.runRunninganmtc.modelusinganMCMCsampler Description Thefunctionmtc.runisusedtogeneratesamplesfromaobjectoftypemtc.modelusingaMCMCsampler.Theresultingmtc.resultobjectcanbecoercedtoanmcmc.listforfurtheranalysisofthedatasetusingthecodapackage.Usage mtc.run(model,sampler=NA,n.adapt=5000,n.iter=20000,thin=1)##S3methodforclassmtc.resultsummary(object,...) ##S3methodforclassmtc.resultplot(x,...) ##S3methodforclassmtc.resultforest(x,use.description=FALSE,...)##S3methodforclassmtc.resultprint(x,...) ##S3methodforclassmtc.resultas.mcmc.list(x,...)Arguments modelsampler AnobjectofS3classmtc.modeldescribinganetworkmeta-analysismodel.Deprecated:gemtcnowonlysupportstheJAGSsampler.Specifyingasam-plerwillresultinawarningorerror.Thisargumentwillberemovedinfutureversions. Amountofadaptation(ortuning)iterations.Amountofsimulationiterations.Thinningfactor. ObjectofS3classmtc.result. n.adaptn.iterthinobject xObjectofS3classmtc.result.use.description DisplaytreatmentdescriptionsinsteadoftreatmentIDs....Value Anobjectofclassmtc.result.Thisisalistwiththefollowingelements:samplesmodel ThesamplesresultingfromrunningtheMCMCmodel,inmcmc.listformat.Themtc.modelusedtoproducethesamples.Additionalarguments. mtc.run deviance 27 Residualdeviancestatistics,alistwiththefollowingelements.DIC:devianceinformationcriterionatresidual(Dbar+pD).Dbar:meansumofresidualde-viance.pD:sumofleverage,alsoknownastheeffectivenumberofparam-eters.dev.ab:meanposteriorresidualdevianceofeacharm(forarm-baseddata).fit.ab:devianceattheposteriormeanoftefittedvalues(forarm-baseddata).dev.re:meanposteriorresidualdevianceofeachstudy(forrelative-effectdata).fit.re:devianceattheposteriormeanofthefittedvalues(forrelative-effectdata). Theobjectcanbecoercedtoanmcmc.listfromthecodapackagebythegenericS3methodas.mcmc.list.Analysisoftheresults Convergenceofthemodelcanbeassessedusingmethodsfromthecodapackage.ForexampletheBrooks-Gelman-Rubinmethod(coda::gelman.diag,coda::gelman.plot).Thesummaryalsoprovidesusefulinformation,suchastheMCMCerrorandthetimeseriesanddensitiesgivenbyplotshouldalsobeinspected. Theforestfunctioncanprovideforestplotsformtc.resultobjects.Thisisespeciallyusefulincombinationwiththerelative.effectfunctionthatcanbeusedtocalculaterelativeeffectscomparedtoanybaselineforconsistencymodels.Therank.probabilityfunctioncalculatesrankprobabilitiesforconsistencymodels.Author(s) GertvanValkenhoef,JoëlKuiperSeeAlso mtc.model relative.effect.table,relative.effect,rank.probabilitycoda::gelman.diag,coda::gelman.plotExamples model<-mtc.model(smoking) ##Notrun:results<-mtc.run(model,thin=10) results<-dget(system.file(\"extdata/luades-smoking.samples.gz\package=\"gemtc\"))#Convergencediagnosticsgelman.plot(results) #Posteriorsummariessummary(results) ##Iterations=5010:25000##Thinninginterval=10##Numberofchains=4 ##Samplesizeperchain=2000## 28 ################################ 1.Empiricalmeanandstandarddeviationforeachvariable, plusstandarderrorofthemean:Mean d.A.B0.4965d.A.C0.8359d.A.D1.1088sd.d0.8465 SD0.40810.24330.43550.1913 NaiveSETime-seriesSE0.0045630.00490.0027200.0031470.0048690.0052800.0021390.002965 parkinson 2.Quantilesforeachvariable:2.5% d.A.B-0.2985d.A.C0.3878d.A.D0.2692sd.d0.5509 25% 0.23120.67200.81970.7119 50%0.49100.82731.09830.8180 75%0.75300.98671.38240.92 97.5%1.3411.3532.0061.283 plot(results)#Showstime-seriesanddensityplotsofthesamplesforest(results)#Showsaforestplot parkinsonMeanoff-timereductioninParkinson’sdisease Description AdatasetofseventrialsinvestigatingfourtreatmentsandplaceboforParkinson’sdisease.Theoutcomeismeanoff-timereduction.Format •parkinson:Anetworkmeta-analysisdatasetcontainingfifteenrowsofarm-baseddata(mean,standarddeviation,andsamplesize). •parkinson_diff:Anetworkmeta-analysisdatasetcontainingfifteenrowsofcontrast-baseddata. •parkinson_shared:Anetworkmeta-analysisdatasetcontainingmixedarm-basedandcontrast-baseddata. Source Franchinietal.(2012),Accountingforcorrelationinnetworkmeta-analysiswithmulti-armtrials,ResearchSynthesisMethods,3(2):142-160.[doi:10.1002/jrsm.1049] plotCovariateEffect29 plotCovariateEffectPlottreatmenteffectsversuscovariatevalues Description Theplotwillshowthemediantreatmenteffectandthe95%credibleintervalonthey-axisandthecovariatevalueonthex-axis.Oneplotpagewillbegeneratedperpairoftreatments.Usage plotCovariateEffect(result,t1,t2,xlim=NULL,ylim=NULL, ask=dev.interactive(orNone=TRUE))Arguments resultt1t2 Resultsobject-createdbymtc.result Alistofbaselinetreatmentstocalculatetreatmenteffectsagainst.Willbeex-tendedtomatchthelengthoft2. Alistoftreatmentstocalculatetheeffectsfor.Willbeextendedtomatchthelengthoft1.Ifleftemptyandt1isasingletreatment,effectsofalltreatmentsexceptt1willbecalculated.Thex-axislimits.They-axislimits. IfTRUE,apromptwillbedisplayedbeforegeneratingthenextpageofamulti-pageplot. xlimylimask Details Defaultx-axislimitswillbesettothreestandarddeviationsaboveandbelowthecenteringvalueofthecovariate.They-axislimitswillbesetbasedontheminimumandmaximum95%CrIlimitsamongthesetofeffectscomputed.Value None.Author(s) GertvanValkenhoefSeeAlso relative.effect,mtc.run 30rank.probability rank.probabilityCalculatingrank-probabilities Description Rankprobabilitiesindicatetheprobabilityforeachtreatmenttobebest,secondbest,etc.Usage rank.probability(result,preferredDirection=1,covariate=NA)##S3methodforclassmtc.rank.probabilityprint(x,...) ##S3methodforclassmtc.rank.probabilityplot(x,...)Arguments result ObjectofS3classmtc.resulttobeusedincreationoftherankprobabilitytable preferredDirection Preferentialdirectionoftheoutcome.Set1ifhighervaluesarepreferred,-1iflowervaluesarepreferred.covariatex...Details ForeachMCMCiteration,thetreatmentsarerankedbytheireffectrelativetoanarbitrarybaseline.Afrequencytableisconstructedfromtheserankingsandnormalizedbythenumberofiterationstogivetherankprobabilities.Value Amatrix(withclassmtc.rank.probability)withthetreatmentsasrowsandtheranksascolumns.Author(s) GertvanValkenhoef,JoëlKuiperSeeAlso relative.effect (Regressionanalysesonly)Valueofthecovariateatwhichtocomputerankprobabilities. AnobjectofS3classrank.probability.Additionalarguments. read.mtc.networkExamples model<-mtc.model(smoking) #Tosavecomputationtimeweloadthesamplesinsteadofrunningthemodel##Notrun:results<-mtc.run(model) results<-dget(system.file(\"extdata/luades-smoking.samples.gz\package=\"gemtc\"))ranks<-rank.probability(results)print(ranks) ##Rankprobability;preferreddirection=1##[,1][,2][,3][,4]##A0.0000000.0030000.1051250.1875##B0.0578750.1758750.6615000.104750##C0.2282500.6005000.1708750.000375##D0.7138750.2206250.0625000.003000plot(ranks)#plotacumulativerankplot plot(ranks,beside=TRUE)#plotarankogram 31 read.mtc.networkReadmtc.networkfromlegacyGeMTCXMLfiles(.gemtc) Description ReadlegacyGeMTCXML(.gemtc)files.Usage read.mtc.network(file)Arguments fileDetails read.mtc.networkexiststomigratedatafromtheolderGeMTCXMLformat,andrequirestheXMLpackagetobeinstalled. SupportforwritingXMLfileshasbeenremoved.Instead,usenativeRmethodstoreadandwritenetworks(e.g.save/loadordput/dget).Value read.mtc.networkreturnsanobjectofS3classmtc.network.Author(s) GertvanValkenhoef Pathtothefiletoread(read.mtc.network). 32SeeAlso mtc.networkExamples #ReadanexampleGeMTCXMLfile file<-system.file(\"extdata/luades-smoking.gemtc\package=\"gemtc\")network<-read.mtc.network(file) #Summarizethenetwork(generatesomeinterestingnetworkproperties)summary(network)##$Description ##[1]\"MTCdataset:Smokingcessationrates\"## ##$Studiespertreatment##ABCD##196196## ##$Numberofn-armstudies##2-arm3-arm##222 relative.effect relative.effectCalculatingrelativeeffects Description Calculatestherelativeeffectsofpairsoftreatments.Usage relative.effect(result,t1,t2=c(),preserve.extra=TRUE,covariate=NA)Arguments resultt1t2 AnobjectofS3classmtc.resulttoderivetherelativeeffectsfrom. Alistofbaselinestocalculatearelativeeffectsagainst.Willbeextendedto matchthelengthoft2. Alistoftreatmentstocalculatetherelativeeffectsfor.Willbeextendedtomatchthelengthoft1.Ifleftemptyandt1isasingletreatment,relativeeffectsofalltreatmentsexceptt1willbecalculated. (Regressionanalysesonly)Valueofthecovariateatwhichtocomputerelativeeffects. preserve.extraIndicateswhethertopreserveextraparameterssuchasthesd.d.covariate relative.effectValue Returnsanmtc.resultsobjectcontainingthecalculatedrelativeeffects. 33 Notethatthismethodstorestherawsamples,whichmayresultinexcessivememoryusage.Youmaywanttoconsiderusingrelative.effect.tableinstead.Author(s) GertvanValkenhoef,JoëlKuiperSeeAlso rank.probability,relative.effect.tableExamples model<-mtc.model(smoking) #Tosavecomputationtimeweloadthesamplesinsteadofrunningthemodel##Notrun:results<-mtc.run(model) results<-dget(system.file(\"extdata/luades-smoking.samples.gz\package=\"gemtc\"))#Createsaforestplotoftherelativeeffectsforest(relative.effect(results,\"A\")) summary(relative.effect(results,\"B\c(\"A\\"C\\"D\")))##Iterations=5010:25000##Thinninginterval=10##Numberofchains=4 ##Samplesizeperchain=2000## ##1.Empiricalmeanandstandarddeviationforeachvariable,##plusstandarderrorofthemean:####MeanSDNaiveSETime-seriesSE##d.B.A-0.49650.40810.0045630.0049##d.B.C0.33940.41440.0046340.004859##d.B.D0.61230.470.00530.005297##sd.d0.84650.19130.0021390.002965## ##2.Quantilesforeachvariable:####2.5%25%50%75%97.5%##d.B.A-1.3407-0.7530-0.4910-0.23120.2985##d.B.C-0.48090.07440.34110.59771.1702##d.B.D-0.30830.30050.60440.91521.5790##sd.d0.55090.71190.81800.921.2827 34relative.effect.table relative.effect.tableTableofrelativeeffects Description Generatesatableoftherelativeeffectsofallpairsoftreatments.Unlikerelative.effect,thismethodstoressummariesonly,notrawsamples.Usage relative.effect.table(result,covariate=NA)##S3methodforclassmtc.relative.effect.tableprint(x,...) ##S3methodforclassmtc.relative.effect.tableforest(x,t1,use.description=FALSE,...) ##S3methodforclassmtc.relative.effect.tableas.data.frame(x,...)Arguments resultcovariatex AnobjectofS3classmtc.resulttoderivetherelativeeffectsfrom. (Regressionanalysesonly)Valueofthecovariateatwhichtocomputerelativeeffects. AnobjectofS3classmtc.relative.effect.table. t1BaselinetreatmentfortheForestplot.use.description DisplaytreatmentdescriptionsinsteadoftreatmentIDs....Value Returnsanmtc.relative.effect.tableobjectcontainingthequantilesofthecalculatedrelativeeffectsofallpair-wisecomparisonsamongthetreatments. Theresultwillbeprettyprintedasann-by-ntableofrelativetreatmenteffects.ItcanalsobeusedtoproduceForestplotsagainstanyarbitrarybaseline.Finally,theas.data.framegenericmethodmakesitpossibletoexportthetableforuseinExcelorotherspreadsheetsoftware,usingthecoreRmethodswrite.csvorwrite.csv2.Author(s) GertvanValkenhoefSeeAlso relative.effect Additionalarguments. smokingExamples model<-mtc.model(smoking) #Tosavecomputationtimeweloadthesamplesinsteadofrunningthemodel##Notrun:results<-mtc.run(model) results<-dget(system.file(\"extdata/luades-smoking.samples.gz\package=\"gemtc\"))#Createsaforestplotoftherelativeeffectstbl<-relative.effect.table(results)#Printthen*ntableprint(tbl) #Ploteffectrelativetotreatment\"C\"forest(tbl,\"C\") #WritetoCSV(e.g.toimporttoExcel,thenuseinaWordtable)write.csv(tbl,\"smoking-effects.csv\") #Note:usewrite.csv2forWesternEuropeanlocales 35 smokingPsychologicaltreatmentstoaidsmokingcessation Description Adatasetof24trialsinvestigatingfourpsychologicaltreatmentsandnotreatmentforsmokingcessation. Format Anetworkmeta-analysisdatasetcontaining50rowsofarm-baseddata(respondersandsamplesize). Source LuandAdes(2006),AssessingEvidenceInconsistencyinMixedTreatmentComparisons,JournaloftheAmericanStatisticalSociety,101(474):447-459.[doi:10.1198/016214505000001302]Hasselblad(1998),Meta-analysisofmultitreatmentstudies,MedicalDecisionMaking18(1):37-43.[doi:10.1177/02729X9801800110] 36thrombolytic thrombolyticThrombolytictreatmentafteracutemyocardialinfarction Description Adatasetof28trialsinvestigatingeightthrombolytictreatmentsadministeredafteramyocardialinfarction.Theoutcomeismortalityafter30-35days.Format Anetworkmeta-analysisdatasetcontaining58rowsofarm-baseddata(respondersandsamplesize).Source LuandAdes(2006),AssessingEvidenceInconsistencyinMixedTreatmentComparisons,JournaloftheAmericanStatisticalSociety,101(474):447-459.[10.1198/016214505000001302] Bolandetal.(2003),Earlythrombolysisforthetreatmentofacutemyocardialinfarction:asystem-aticreviewandeconomicevaluation,HealthTechnologyAssessment7(15):1-136.[doi:10.3310/hta7150] Index [Diaset[Diasetal.al.2010,vanValkenhoefet al.2015]2013a,,19 vanValkenhoefet [Dias[Diasetal.etal.2012]2013a],19 ,19 [Turneral.al.(draft)]2013b,van,Valkenhoefet [Warn[van[vanValkenhoefetetal.al.2002]2012],19 17Valkenhoefet,19 etal.al.(draft)]2014b(draft)],19, 13 [vanValkenhoefetal.2015],24as.data.frame.mtc.relative.effect.table (relative.effect.table),34 as.mcmc.list.mtc.resultatrialFibrillation,4,20(mtc.run),26blobbogramblocker,8,5certolizumab,8,20depressiondietfat,10 ,9forestforest.mtc.relative.effect.table (blobbogram),5 (relative.effect.table),34 forest.mtc.result(mtc.run),26Gelmangemtcgemtc-package(gemtc-packageetal.(2008),2),,219hfPrevention,10,20ll.call,11 mtc(gemtc-package),2 37 mtc.anohemtc.data.studyrow,12 mtc.deviancemtc.devplot,15,16 ,14,23mtc.hy.empirical.lor(mtc.deviance),15 mtc.hy.priormtc.levplot,(mtc.hy.prior),16mtc.modelmtc.network,3,(12mtc.deviance16,18 ,13,16,17,18),15 ,22–25,27mtc.nodesplit,3,13–15,21,22,24,mtc.resultmtc.result,16 ,2432mtc.run,3,13(mtc.run,16,21,),2426 ,25,26,29parkinsonparkinson_diff,28 parkinson_shared(parkinsonplot.mtc.anohe(parkinson),28plot.mtc.deviance(mtc.anoheplot.mtc.model(mtc.deviance),12 ),28),plot.mtc.network(mtc.model),18 15plot.mtc.nodesplit(mtc.network),22plot.mtc.rank.probability (mtc.nodesplit),24(rank.probability),30 plot.mtc.resultplotCovariateEffect(mtc.run),26print.mtc.anohe,29 print.mtc.model(print.mtc.nodesplit(mtc.anohemtc.model),12print.mtc.rank.probability (mtc.nodesplit),18 ),24(rank.probability),30 print.mtc.relative.effect.table (relative.effect.table),34 print.mtc.result(mtc.run),26rank.probabilityread.mtc.network,27,30,33relative.effectrelative.effect.table,,2731 ,29,,3027,,3233,,3434 38 smoking,35 summary.mtc.anohe(mtc.anohe),12summary.mtc.model(mtc.model),18 summary.mtc.nodesplit(mtc.nodesplit), 24 summary.mtc.result(mtc.run),26thrombolytic,36 INDEX 因篇幅问题不能全部显示，请点此查看更多更全内容
Copyright © 2019- igat.cn 版权所有 赣ICP备2024042791号-1
违法及侵权请联系：TEL:199 1889 7713 E-MAIL:2724546146@qq.com
本站由北京市万商天勤律师事务所王兴未律师提供法律服务