资源描述:
《a statistical explanation英文学习材料》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、DiversityandDistributions,(DiversityDistrib.)(2011)17,43–57BIODIVERSITYAstatisticalexplanationofMaxEntforRESEARCHecologists1234JaneElith*,StevenJ.Phillips,TrevorHastie,MiroslavDudı´k,15YungEnCheeandColinJ.Yates1SchoolofBotany,TheUniversityofABSTRACTMelbourne,Parkville,VIC3010Australia,2Ma
2、xEntisaprogramformodellingspeciesdistributionsfrompresence-onlyAT&TLabsResearch,180ParkAvenue,3speciesrecords.ThispaperiswrittenforecologistsanddescribestheMaxEntFlorhamPark,NJ07932,USA,Departmentmodelfromastatisticalperspective,makingexplicitlinksbetweenthestructureofofStatistics,Stanfor
3、dUniversity,CA94305,4themodel,decisionsrequiredinproducingamodelleddistribution,andUSA,Yahoo!Labs,111West40thStreet(17thFloor).NewYork,NY10018,USA,knowledgeaboutthespeciesandthedatathatmightaffectthosedecisions.To5ScienceDivision,WesternAustralianbeginwediscussthecharacteristicsofpresence
4、-onlydata,highlightingimplica-DepartmentofEnvironmentandtionsformodellingdistributions.WeparticularlyfocusontheproblemsofConservation,LMB104,BentleyDeliverysamplebiasandlackofinformationonspeciesprevalence.ThekeystoneoftheCentre,WA6983,AustraliapaperisanewstatisticalexplanationofMaxEntwhi
5、chshowsthatthemodelAJournalofConservationBiogeographyminimizestherelativeentropybetweentwoprobabilitydensities(oneestimatedfromthepresencedataandone,fromthelandscape)definedincovariatespace.Formanyusers,thisviewpointislikelytobeamoreaccessiblewaytounderstandthemodelthanpreviousonesthatrely
6、onmachinelearningconcepts.WethenstepthroughadetailedexplanationofMaxEntdescribingkeycomponents(e.g.covariatesandfeatures,anddefinitionofthelandscapeextent),themechanicsofmodelfitting(e.g.featureselection,constraintsandregularization)andoutputs.UsingcasestudiesforaBanksiaspeciesnativetosouth
7、-westAustraliaandariverinefish,wefitmodelsandinterpretthem,exploringwhycertainchoicesaffecttheresultandwhatthismeans.Thefishexampleillustratesuseofthemodelwithvectordataforlinearriversegmentsratherthanraster(gridded)data.Appropriatetreatmentsforsurveybias,unprojectedda