GWR4-Manual-软件GWR4操作指南

GWR4UserManualGWR4WindowsApplicationforGeographicallyWeightedRegressionModellingTomokiNakayaUpdated7May20123June2009GWR4DevelopmentTeamTomokiNakaya(DepartmentofGeography,RitsumeikanUniversity)MartinCharlton,PaulLewis(NCG,NationalUniversityofIreland)StewartFotheringham(SchoolofGeography&Geosciences,UniversityofSt.Andrews)ChrisBrunsdon(DepartmentofGeography,UniversityofLiverpool)1Contents1.Introduction...................................................................................................2WhatisGWR4?....................................................................................................2NewfeaturescomparedtoGWR3.x....................................................................3NotesforuseofGWR4........................................................................................42.Installation/Uninstallation..........................................................................43.Startingtheprogram,Exitingtheprogram,andTabdesign.......................5FivestepsinGWRcalibration............................................................................64.Step1:TheDataTab.....................................................................................7Datapreparation.................................................................................................7Operationsinthedatatabpage.........................................................................95.Step2:TheModelTab.................................................................................10Basicoperations:usinganexampleofGaussianGWR...................................11SemiparametricGWR........................................................................................13ExtensionsofGWR:GWGLM...........................................................................15GeographicallyweightedPoissonregression(GWPR).................................15Geographicallyweightedlogisticregression(GWLR)..................................16ModellingOptions..............................................................................................18Standardisation..............................................................................................18Geographicalvariabilitytest.........................................................................18LtoG/GtoLvariableselection.......................................................................206.Step3:TheKernelTab................................................................................22Possiblefixedandadaptivekernelfunctionsforgeographicalweighting......23Bandwidthselectionroutines............................................................................24Selectioncriteria................................................................................................267.Step4:TheOutputTab...............................................................................27SessionControlFile...........................................................................................27Commonoutputfiles.........................................................................................28The“Predictionatnon-regressionpoints”option............................................288.Step5:TheExecuteTab..............................................................................29TheExecutebutton............................................................................................29Fieldsinalistwiseoutput.................................................................................31Handlingasessionandbatchmode(optional)................................................32Exampleoutput.................................................................................................339.References....................................................................................................3921.IntroductionWhatisGWR4?GWR4isanewreleaseofaMicrosoftWindows-basedapplicationsoftwareforcalibratinggeographicallyweightedregression(GWR)models,whichcanbeusedtoexploregeographicallyvaryingrelationshipsbetweendependent/responsevariablesandindependent/explanatoryvariables.AGWRmodelcanbeconsideredatypeofregressionmodelwithgeographicallyvaryingparameters.AconventionalGWRisdescribedbytheequationkiikiikixvuy,),(,whereiy,ikx,,andiare,respectively,dependentvariable,kthindependentvariable,andtheGaussianerroratlocationi;(iivu,)isthex-ycoordinateoftheithlocation;andcoefficients),(iikvuarevaryingconditionalsonthelocation.Suchmodellingislikelytoattainhigherperformancethantraditionalregressionmodels,andreadingthecoefficientscanleadtoanewinterpretationofthephenomenaunderstudy.AnimportantextensionofGWRisitssemiparametricformationbymixinggloballyfixedandgeographicallyvaryingcoefficients.kilillikiikizxvuy,,),(.whereilz,isthelthindependentvariablewithafixedcoefficientl.Suchamodelmayreducethemodelcomplexitiesandenhanceitspredictableperformance.Usingtheframeworkofgeographicallyweightedgeneralisedlinearmodelling(GWGLM),logisticandPoissonregressionmodelswithgeographic