Operational Risk Modelling and Management 3

《Operational Risk Modelling and Management 3》由会员上传分享，免费在线阅读，更多相关内容在学术论文-天天文库。

1、Chapter2TheProblemContext2.1GeneralRemarksModellingalwaysmeansreducingtheobjectunderinvestigationtoitsmainrelevantfeatures.Itisaprocessofoptimisationthatgaugesthecomplexityandtractabilityoftheobjectinquestionusinganagreed-uponmethodology.Inthisinstancewemustcons

2、ideringmathematicalmodelswhose“mainfeatures”arerepresentedbylogicalandquantitativerelationships.Modelsareusefulbecausethey(Fishman,1978):•enableinvestigatorstoorganizetheoreticalbeliefsandempiricalobser-vationsaboutasystemandtodeducethelogicalimplicationsofthiso

3、rganisation,•leadtoimprovedOperationalRiskManagementsystemunderstanding,•bringintoperspectivetheneedfordetailandrelevanceand•provideaframeworkfortestingthedesirabilityofOperationalRiskManagementmodifications.Itisevidentthatseveralmodelscanco-exist.Wepresentjuston

4、emodelthat:•complieswiththerestrictionsoftheBasel2framework,•isbasedonlong-standingactuarialmethodologyand,•triestobalanceavailabledataanddatamodelling.Thereisanintimaterelationshipbetweendataandmodelwhichisanalogoustothelogicofreasoning(seeFigure2.1).Thestructu

5、reconsistsofinductiveanddeductiveinferences.Inthefirstcase,weinferfromthespecifictothegeneralbydrawingconclusionsfromasetofpremises,suchasexperience,ev-idenceanddata.Theconclusionscontainmoreinformationthanthepremisesandmaybedisprovedbyfurtherevidence.Inthecaseofd

6、eductiveinferences,thereverseistrue:wemovefromthegeneraltothespecificbydrawingconclu-sionsthatnecessarilyfollowfromthegeneralsetofpremises.Inourcontext,inductionisassociatedwithStatisticswhiledeductionpertainstoProbability49©2011byTaylorandFrancisGroup,LLC50Opera

7、tionalRiskModellingandManagementInductionSpecificStatisticsGeneralSamplePopulationDataDeductionModelProbabilityFIGURE2.1:Dataandmodel.theory.Asourmodelisbasedupongivenandaccumulatedlossdata,itmaybethatfuturedatawillnecessitatearevisionofourmodel.ThereforetheLDAw

8、ithitsactualpremisesmustbeconsideredatemporaryphenomenon.2.2TheDataProblemThedataavailableformodellinglosseswithadistributionapproachisboundbyquantityandquality.Formo