ebook img

Risk Analysis of Complex and Uncertain Systems PDF

453 Pages·2009·2.57 MB·English
Save to my drive
Quick download
Download
Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.

Preview Risk Analysis of Complex and Uncertain Systems

Risk Analysis of Complex and Uncertain Systems INT.SERIESINOPERATIONSRESEARCH&MANAGEMENTSCIENCE SeriesEditor:FrederickS.Hillier,StanfordUniversity SpecialEditorialConsultant:CamilleC.Price,StephenF.AustinStateUniversity Titleswithanasterisk(∗)wererecommendedbyDr.Price Axsa¨ter/INVENTORYCONTROL,2ndEd. Hall/PATIENTFLOW:ReducingDelayinHealthcareDelivery Jo´zefowska&We˛glarz/PERSPECTIVESINMODERNPROJECTSCHEDULING Tian&Zhang/VACATIONQUEUEINGMODELS:TheoryandApplications Yan,Yin&Zhang/STOCHASTICPROCESSES,OPTIMIZATION,ANDCONTROLTHEORY APPLICATIONSINFINANCIALENGINEERING,QUEUEINGNETWORKS,AND MANUFACTURINGSYSTEMS Saaty&Vargas/DECISIONMAKINGWITHTHEANALYTICNETWORKPROCESS:Economic, Political,Social&TechnologicalApplicationswithBenefits,Opportunities,Costs&Risks Yu/TECHNOLOGYPORTFOLIOPLANNINGANDMANAGEMENT:PracticalConceptsandTools Kandiller/PRINCIPLESOFMATHEMATICSINOPERATIONSRESEARCH Lee&Lee/BUILDINGSUPPLYCHAINEXCELLENCEINEMERGINGECONOMIES Weintraub/MANAGEMENTOFNATURALRESOURCES:AHandbookofOperationsResearch Models,Algorithms,andImplementations Hooker/INTEGRATEDMETHODSFOROPTIMIZATION Dawandeetal./THROUGHPUTOPTIMIZATIONINROBOTICCELLS Friesz/NETWORKSCIENCE,NONLINEARSCIENCE,andINFRASTRUCTURESYSTEMS Cai,Sha&Wong/TIME-VARYINGNETWORKOPTIMIZATION Mamon&Elliott/HIDDENMARKOVMODELSINFINANCE delCastillo/PROCESSOPTIMIZATION:AStatisticalApproach Jo´zefowska/JUST-IN-TIMESCHEDULING:Models&AlgorithmsforComputer&Manufacturing Systems Yu,Wang&Lai/FOREIGN-EXCHANGE-RATEFORECASTINGWITHARTIFICIALNEURAL NETWORKS Beyeretal./MARKOVIANDEMANDINVENTORYMODELS Shi&Olafsson/NESTEDPARTITIONSOPTIMIZATION:MethodologyandApplications Samaniego/SYSTEMSIGNATURESANDTHEIRAPPLICATIONSINENGINEERINGRELIABILITY Kleijnen/DESIGNANDANALYSISOFSIMULATIONEXPERIMENTS Førsund/HYDROPOWERECONOMICS Kogan&Tapiero/SUPPLYCHAINGAMES:OperationsManagementandRiskValuation Vanderbei/LINEARPROGRAMMING:Foundations&Extensions,3rdEdition Chhajed&Lowe/BUILDINGINTUITION:InsightsfromBasicOperationsMgmt.Modelsand Principles Luenberger&Ye/LINEARANDNONLINEARPROGRAMMING,3rdEdition Drewetal./COMPUTATIONALPROBABILITY:AlgorithmsandApplicationsintheMathematical Sciences∗ Chinneck/FEASIBILITYANDINFEASIBILITYINOPTIMIZATION:AlgorithmsandComputation Methods Tang,Teo&Wei/SUPPLYCHAINANALYSIS:AHandbookontheInteractionofInformation, System,andOptimization Ozcan/HEALTHCAREBENCHMARKINGANDPERFORMANCEEVALUATION:AnAssessment UsingDataEnvelopmentAnalysis(DEA) Wierenga/HANDBOOKOFMARKETINGDECISIONMODELS Agrawal&Smith/RETAILSUPPLYCHAINMANAGEMENT:QuantitativeModelsandEmpirical Studies ∼Alistoftheearlypublicationsintheseriesisfoundattheendofthebook∼ Louis Anthony Cox, Jr. Risk Analysis of Complex and Uncertain Systems 123 LouisAnthonyCox,Jr. CoxAssociates 503FranklinStreet DenverCO80218 USA [email protected] ISBN978-0-387-89013-5 e-ISBN978-0-387-89014-2 DOI10.1007/978-0-387-89014-2 LibraryofCongressControlNumber:2008940639 (cid:2)c SpringerScience+BusinessMedia,LLC2009 Allrightsreserved.Thisworkmaynotbetranslatedorcopiedinwholeorinpartwithoutthewritten permission of the publisher (Springer Science+Business Media, LLC, 233 Spring Street, New York, NY10013,USA),exceptforbriefexcerptsinconnectionwithreviewsorscholarlyanalysis.Usein connection with any form of information storage and retrieval, electronic adaptation, computer software,orbysimilarordissimilarmethodologynowknownorhereafterdevelopedisforbidden. The use in this publication of trade names, trademarks, service marks, and similar terms, even if they are not identified as such, is not to be taken as an expression of opinion as to whether or not theyaresubjecttoproprietaryrights. Printedonacid-freepaper springer.com ToChristineandEmeline Preface WhyThisBook? Thisbookismotivatedbythefollowingconvictions: 1) Quantitativeriskassessment(QRA)canbeapowerfuldisciplineforimproving riskmanagementdecisionsandpolicies. 2) Poorly conducted QRAs can produce results and recommendations that are worsethanuseless. 3) Sound riskassessmentmethods provide thebenefits ofQRAmodeling –being able to predict and compare the probable consequences of alternative actions, interventions, or policies and being able to identify those that make preferred consequencesmoreprobable–whileavoidingthepitfalls. ThisbookdevelopsandillustratesQRAmethodsforcomplexanduncertainbio- logical,engineering,andsocialsystems.Thesesystemshavebehaviorsthataretoo complexoruncertaintobemodeledaccuratelyindetailwithhighconfidence.Prac- ticalapplicationsincludeassessingandmanagingrisksfromchemicalcarcinogens, antibioticresistance,madcowdisease,terroristattacks,andaccidentalordeliberate failuresintelecommunicationsnetworkinfrastructure. ForWhomIsItMeant? This book is intended primarily for practitioners who want to use rational quanti- tativeriskanalysistosupportandimproveriskmanagementdecisionsinimportant health, safety, environmental, reliability, and security applications, but who have beenfrustratedintryingtoapplytraditionalquantitativemodelingmethodsbythe highuncertaintyand/orcomplexityofthesystemsinvolved.Weemphasizemethods andstrategiesformodelingcausalrelationsincomplexanduncertainsystemswell enoughtomakeeffectiveriskmanagementdecisions.Thebookiswrittenforpracti- tionersfrommultipledisciplines–decisionandriskanalysts,operationsresearchers and management scientists, quantitative policy analysts, economists, health and vii viii Preface safetyriskassessors,engineers,andmodelers–whoneedpracticalwaystopredict andmanagerisksincomplexanduncertainsystems. What’sinIt? Three introductory chapters describe QRA and compare it to less formal alterna- tives, such as taking prompt action to address current concerns, even if the con- sequences caused by the recommended action are unknown (Chapter 1). These chapters survey QRA methods for engineering risks (Chapter 2) and health risks (Chapter3).Briefexamplesofapplicationssuchasfloodcontrol,softwarefailures, chemicalreleases,andfoodsafetyillustratethescopeandcapabilitiesofQRAfor complexanduncertainsystems. Chapter 1 discusses a concept of concern-driven risk management, in which qualitative expert judgments about whether concerns warrant specified risk man- agementinterventionsareusedinpreferencetoQRAtoguideriskmanagementde- cisions.WhereQRAemphasizestheformalquantitativeassessmentandcomparison oftheprobableconsequencescausedbyrecommendedactionstotheprobablecon- sequencesofalternatives,includingthestatusquo,concern-drivenriskmanagement insteademphasizestheperceivedurgencyorseverityofthesituationmotivatingrec- ommendedinterventions.Inmanyinstances,especiallythoseinvolvingapplications of a “Precautionary Principle” (popular in much European legislation), no formal quantification or comparison of probable consequences for alternative decisions is seen as being necessary (or, perhaps, possible or desirable) before implementing riskmanagementmeasuresthatareintendedtopreventseriousorirreversibleharm, evenifthecausalrelationsbetweentherecommendedmeasuresandtheirprobable consequences are unclear. Such concern-driven risk management has been recom- mendedbycriticsofQRAinseveralareasofappliedriskmanagement. Basedoncasestudiesandpsychologicalliteratureontheempiricalperformance of judgment-based decision making under risk and uncertainty, we conclude that, althoughconcern-drivenriskmanagementhasseveralimportantpotentialpolitical and psychological advantages over QRA, it often performs less well than QRA in identifyingriskmanagementinterventionsthatsuccessfullyprotecthumanhealthor achieveotherdesiredconsequences.Therefore,thosewhoadvocatereplacingQRA with concern-driven alternatives, such as expert judgment and consensus decision processes, should assess whether their recommended alternatives truly outperform QRA, by the criterion of producing preferred consequences, before rejecting the QRAparadigmforpracticalapplications. Chapter2introducesmethodsofprobabilisticriskassessment(PRA)forpredict- ingandmanagingrisksincomplexengineeredsystems.ItsurveysmethodsforPRA and decision making in engineered systems, emphasizing progress in methods for dealingwithuncertainties,communicatingresultseffectively,andusingtheresults toguideimproveddecisionmakingbymultipleparties.Forsystemsoperatingunder threats from intelligent adversaries, novel methods and game-theoretic ideas can Preface ix helptoidentifyeffectiveriskreductionstrategiesandresourceallocations.Inhard decision problems, where the best course of action is unclear and data are sparse, ambiguous,orconflicting,state-of-the-artmethodologycanbecriticalforgoodrisk management. This chapter discusses some of the most useful PRA methods and possibleextensionsandimprovements. Chapter 3 introduces methods of quantitative risk assessment (QRA) for pub- lichealthrisks.Thesearisefromtheoperationofcomplexengineering,economic, medical,andsocialsystems,rangingfromfoodsupplynetworkstoindustrialplants toadministrationofschoolvaccinationprogramsandhospitalinfectioncontrolpro- grams.Thedecisionsandbehaviorsofmultipleeconomicagents(e.g.,theproduc- ers, distributors, retailers, and consumers of a product) or other decision makers (e.g.,parents,physicians,andschoolsinvolvedinvaccinationprograms)affectrisks that,inturn,typicallyaffectmanyotherpeople.Healthrisksarecommonlydifferent fordifferentsubpopulations(e.g.,infants,theelderly,andtheimmunocompromised, for a microbial hazard; or customers, employees, and neighbors of a production process).Thus,publichealthriskanalysisoftenfallsintheintersectionofpolitics, business,law,economics,ethics,science,andtechnology,withdifferentparticipants andstakeholdersfavoringdifferentriskmanagementalternatives.Inthispoliticized context,QRAseekstoclarifytheprobableconsequencesofdifferentriskmanage- mentdecisions. Chapters4and5(aswellasChapter15,whichdealsspecificallywithterrorism risk assessment) emphasize that sound risk assessment requires developing sound riskmodelsinenoughdetailtorepresentcorrectlythe(oftenprobabilistic)causalre- lationsbetweenasystem’scontrollableinputsandtheoutputsorconsequencesthat decision makers care about. “Sound” does not imply completely accurate, certain, ordetailed.Imperfectandhigh-levelriskmodels,orsetsofalternativeriskmodels that are contingent on explicitly stated assumptions, can still be sound and useful for improving decision making. But a sound model must describe causal relations correctly, even ifnot ingreat detail,and even ifcontingent on stated assumptions. Incorrectcausalmodels,ormodelswithhiddenfalseassumptionsaboutcauseand effect,canleadtopoorriskmanagementrecommendationsanddecisions. Chapters 4 and 5 warn against popular shortcut methods of risk analysis that try to avoid the work required to develop and validate sound risk models. These includereplacingempiricallyestimatedandvalidatedcausalriskmodels(e.g.,sim- ulation models) with much simpler ratings of risky prospects using terms such as high,medium,andlowforattributessuchasthefrequencyandseverityofadverse consequences.Othershortcutmethodsusehighlyaggregateriskmodelsorscoring formulas(suchas“risk=potency×exposure,”or“risk=threat×vulnerability× consequence”) in place of more detailed causal models. Many professional con- sultants,riskassessors,andregulatoryagenciesusesuchmethodstoday.However, theseattemptedshortcutsdonotworkwellingeneral.AsdiscussedinChapters4 and5,theycanproduceresults,recommendations,andprioritiesthatareworsethan useless:theyareevenlesseffective,onaverage,thanmakingdecisionsrandomly! Poorriskmanagementdecisions,basedonfalsepredictionsandassumptions,result fromtheseshortcutmethods. x Preface Fortunately,itispossibletodomuchbetter.Buildingandvalidatingsoundcausal riskmodelsleadstoQRAmodelsandanalysesthatcangreatlyimproveriskman- agement decisions. Chapters 6 through 16 explain how. They introduce and illus- trate techniques for testing causal hypotheses and for identifying potential causal relationsfromdata(Chapters6and7),fordeveloping(andempiricallytestingand validating)riskmodelstopredicttheresponsesofcomplex,uncertain,andnonlinear systems to changes in controllable inputs (Chapters 8-13), and for making more effectiveriskmanagementdecisions,despiteuncertaintiesandcomplexities(Chap- ters14-16).Thesechaptersposeavarietyofimportantriskanalysischallengesfor complexanduncertainsystems,andproposeandillustratemethodsforsolvingthem inimportantreal-worldapplications. Keychallenges,methodsandapplicationsinChapters6through16includethe following: (cid:2) Information-theoryanddata-miningalgorithms.Chapter6showshowtodetect initially unknown, possibly nonlinear (including u-shaped) causal relations in epidemiological data sets, using food poisoning data as an example. A combi- nation of information theory and nonparametric modeling methods (especially, classification tree algorithms) provide constructive ways to identify potential causal relations (including nonlinear and multivariate ones with high-order in- (cid:2) teractions)inmultivariateepidemiologicaldatasets. Testingcausalhypothesesanddiscoveringcausalrelations.Chapter7,building onthemethodsinChapter6,discusseshowtotestcausalhypothesesusingdata, how to discover new causal relations directly from data without any a priori hypotheses, and how to use data mining and other statistical methods to avoid imposingone’sownpriorbeliefsontheinterpretationofdata–aperennialchal- lenge in risk assessment and other quantitative modeling disciplines. An appli- (cid:2) cationtoantibiotic-resistantbacterialinfectionsillustratesthesetechniques. Useofnew molecular-biological and “-omics” informationinriskassessment. Chapter 8 shows how to use detailed biological data (arising from advances in genomics,proteomics,metabolomics,andotherlow-levelbiologicaldata)topre- dictthefractionofillnesses,diseases,orotherunwantedeffectsinapopulation that could be prevented by removing specific hazards or sources of exposure. Thischallengeisaddressedbyusingconditionalprobabilityformulasandcon- servative upper bounds on the observed occurrence and co-occurrence rates of events in a causal network to obtain useful upper bounds on unknown causal fractions. Bounding calculations are illustrated by quantifying the preventable fraction of smoking-associated lung cancers in smokers caused by – and pre- ventable by blocking – a particular causal pathway (involving polycyclic aro- matic hydrocarbons forming adducts with DNA in a critical tumor suppressor (cid:2) gene)thathasattractedgreatrecentinterest. Upper-boundingmethods.Chapters8through12considerhowtouseavailable knowledge and information about causal pathways in complex systems, even if very imperfect and incomplete (e.g., biomarker data for complex diseases), to estimate upper bounds on the preventable fractions of disease that could be

Description:
"Tony Cox is among the most active and creative architects and users of quantitative risk analysis. This book is full of interesting equations, conceptual designs and conundrums that characterize QRA and its applications to risk management. Informed by trenchant thinking and perceptive writing, this
See more

The list of books you might like

Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.