nach oben

2009 | Buch

Kapitel lesen Erstes Kapitel lesen

Risk Analysis of Complex and Uncertain Systems

verfasst von: Louis Anthony Cox

Verlag: Springer US

Buchreihe : International Series in Operations Research & Management Science

Enthalten in: Springer Professional "Wirtschaft+Technik" , Springer Professional "Technik" , Springer Professional "Wirtschaft"

Einloggen, um Zugang zu erhalten

Über dieses Buch

In Risk Analysis of Complex and Uncertain Systems acknowledged risk authority Tony Cox shows all risk practitioners how Quantitative Risk Assessment (QRA) can be used to improve risk management decisions and policies. It develops and illustrates QRA methods for complex and uncertain biological, engineering, and social systems – systems that have behaviors that are just too complex to be modeled accurately in detail with high confidence – and shows how they can be applied to applications including assessing and managing risks from chemical carcinogens, antibiotic resistance, mad cow disease, terrorist attacks, and accidental or deliberate failures in telecommunications network infrastructure. This book was written for a broad range of practitioners, including decision risk analysts, operations researchers and management scientists, quantitative policy analysts, economists, health and safety risk assessors, engineers, and modelers.

Inhaltsverzeichnis

Frontmatter

PART 1 INTRODUCTION TO RISK ANALYSIS

Frontmatter

Chapter 1. Quantitative Risk Assessment Goals and Challenges

How should societies, organizations, and individuals manage risks from activities with unknown or uncertain consequences? Many regulators and scientists advocate quantitative risk assessment (QRA) as providing both a logical framework and a systematic procedure for organizing and applying scientific and engineering knowledge to improve “rational” (consequence-driven) decision making when the consequences of alternative decisions are uncertain. It seeks to do so by using predictive models to identify and recommend choices (typically, among alternative risk management interventions, policies, or plans) that are predicted to make preferred consequences more likely.

Louis Anthony Cox Jr

Chapter 2. Introduction to Engineering Risk Analysis

Can contemporary organizations and societies design, build, and operate complex engineering systems safely and reliably for long periods? Being able to do so is crucial if nuclear power is to be a viable option, if infrastructure such as advanced transportation systems and energy distribution networks is to be trustworthy, if minerals and petroleum are to be discovered and extracted safely, and if hazardous manufacturing and chemical storage facilities are to be located in convenient proximity to transportation hubs and population centers. This chapter, which is an update and extension of Bier and Cox (2007), discusses methods for quantifying the extent to which complex engineering systems can be designed and operated safely.

Louis Anthony Cox Jr

Chapter 3. Introduction to Health Risk Analysis

This chapter, which is an update of Cox (2007), introduces methods of quantitative risk assessment (QRA) for public health risks. As noted in the Preface, public health risk analysis often falls in the intersection of politics, business, law, economics, and science and technology, as stakeholders with different interests seek to use QRA for their own ends. Public health risk analysis deals with decisions about which potential risk management interventions (usually including the status quo or “do-nothing” option) should be implemented to maintain or increase the safety of complex social, economic, and technological systems, such as the food supply network or industrial emissions control systems. The best course of action is often hotly disputed. For example, should emissions of gases or particles from a facility be further restricted even if permitting the current levels has unquestioned benefits in industry or agriculture? Should cell phone use in cars be banned? (“Public health” is often extended to include such issues of public safety.) Should cattle be imported from countries that have low levels of diseases such as BSE? Should antibiotics used in human medicine be prohibited from uses in food animals, even if doing so will lead to more sick animals (and perhaps more sick people), in order to preserve the effectiveness of the antibiotics in treating human patients? QRA seeks to inform policy-making deliberation and debate in such controversial cases by clarifying the probable consequences of alternative decisions.

Louis Anthony Cox Jr

PART 2 AVOIDING BAD RISK ANALYSIS

Chapter 4. Limitations of Risk Assessment Using Risk Matrices

This chapter focuses on the use and abuse of risk matrices – tables mapping “frequency” and “severity” ratings to corresponding risk priority levels. Such matrices have become very popular in applications as diverse as terrorism risk analysis, highway construction project management, office building risk analysis, climate change risk management, and enterprise risk management (ERM). Their use is now so widespread in important applications that it is worth devoting an entire chapter to understanding the main concepts – and limitations – of risk matrices as a framework for practical risk analysis.

Louis Anthony Cox Jr

Chapter 5. Limitations of Quantitative Risk Assessment Using Aggregate Exposure and Risk Models

Chapter 4 showed that risk matrices can assign small risks to high-risk categories and larger risks to lower-risk categories, defeating the intent of the classification system. Do other methods necessarily do better? This chapter shows that careless use of quantitative risk assessment concepts can also lead to worse-than-useless risk comparisons and recommendations. This happens if causal drivers of risk (such as age-specific failure rates, detailed exposures, or individual dose-response relations) are ignored in favor of potentially meaningless aggregate quantities (such as “average annual frequency,” “aggregate exposure,” or “population exposure-response ratio,” respectively). A lesson from Chapter 4 was that risk matrices cannot correctly compare some risks. The main lesson from this chapter is milder. Care must be taken in using quantitative risk concepts to make sure that they correctly represent causal relations among actions, exposures, and probable consequences. Otherwise, they may give rise to meaningless or misleading numbers and predictions.

Louis Anthony Cox Jr

PART 3 PRINCIPLES FOR DOING BETTER

Chapter 6. Identifying Nonlinear Causal Relations in Large Data Sets

This chapter discusses data-mining methods for identifying potential causal relations in large data sets, such as clinical, epidemiological, or engineering reliability data sets. The causal relations to be discovered may be completely unknown initially; thus, successfully identifying them from data is sometimes called knowledge discovery. This is usually more challenging than merely estimating the parameters of a statistical model that is known or specified a priori. The causal relations may be complex and impossible to summarize using only a few parameters. For example, they may contain nonmonotonic (such as n-shaped or u-shaped) or threshold-like exposure-response relations, or more complicated nonlinearities, that render ineffective traditional statistical data analysis techniques (including factor analysis, principal components analysis, discriminant analysis, multiple linear or logistic regression, and so forth) based on linear and generalized linear modeling.

Louis Anthony Cox Jr

Chapter 7. Overcoming Preconceptions and Confirmation Biases Using Data Mining

Data-mining methods such as classification tree analysis, conditional independence tests, and causal graphs can be used to discover possible causal relations in data sets, even if the relations are unknown a priori and involve nonlinearities and high-order interactions. Chapter 6 showed that information theory provided one possible common framework and set of principles for applying these methods to support causal inferences. This chapter examines how to apply these methods and related statistical techniques (such as Bayesian model averaging) to empirically test preexisting causal hypotheses, either supporting them by showing that they are consistent with data, or refuting them by showing that they are not. In the latter case, data-mining and modeling methods can also suggest improved causal hypotheses.

Louis Anthony Cox Jr

Chapter 8. Estimating the Fraction of Disease Caused by One Component of a Complex Mixture: Bounds for Lung Cancer

Lung cancer illustrates many of the challenges of modeling cause and effect in very complex systems having only poorly understood causal mechanisms. When not enough is known to develop useful marginal and conditional probability distributions for uncertain quantities, it may be practical instead to develop bounds on uncertain quantities and causal relations. This chapter illustrates bounding for lung cancer risks. Other areas of quantitative modeling and operations research, from robust optimization to constraint logic programming, apply a similar insight: It is often practical to use limited available data to develop bounds on the likely consequences of actions, even if the data are not adequate to estimate informative, well-calibrated, probabilities for consequences. This chapter and Chapter 9 discuss risk bounds for uncertain complex systems. To illustrate how to develop bounds from data, we quantify bounds on preventable disease risks for two very different illnesses – lung cancer and penicillin-resistant bacterial infections, respectively.

Louis Anthony Cox Jr

Chapter 9. Bounding Resistance Risks for Penicillin

Chapter 8 calculated a plausible upper bound for the fraction of disease preventable by blocking a specific causal pathway in a complex, uncertain biological system, using smoking-induced lung cancer as an example. This example required considering relevant biological knowledge and biomarker data in some detail. But plausible upper bounds on preventable risks can also be developed using much less detailed knowledge, and data that are relatively easy to obtain and understand, for many other systems. This chapter develops bounding calculations for preventable fractions in a very different complex uncertain system: the highly uncertain set of pathways and mechanisms (if any) leading from the use of penicillin in food animals to resistance to penicillin drugs in human patients. Unlike the campylobacteriosis and virginiamycin case studies in Chapters 6 and 7, respectively, no individual-level exposure and response data are provided for this penicillin risk assessment.

Louis Anthony Cox Jr

Chapter 10. Confronting Uncertain Causal Mechanisms – Portfolios of Possibilities

This chapter returns to smoking and lung cancer risk as a fruitful example of a complex system with many uncertainties (and, as discussed in Chapter 11, nonlinearities) in its input-output (dose-response) relations. These uncertainties, complexities, and nonlinearities raise important challenges for quantitative risk assessment (QRA) modeling. The challenge confronted in this chapter is how to estimate the potential effects on lung cancer of removing a specific constituent, cadmium (Cd), from cigarette smoke, given the very incomplete scientific information available now about its possible modes of carcinogenic action. Not enough is known about how cadmium affects lung cancer to allow useful bounds on risk to be established using biomarkers, as in Chapter 8. A different strategy is needed for QRA.

Louis Anthony Cox Jr

Chapter 11. Determining What Can Be Predicted: Identifiability

One of the best developed ways to predict how changing inputs to a complex system will change its probable outputs is to simulate the behavior of the system. Modern simulation modeling software environments (such as MATLAB/SIMULINK^®, or STELLA/ITHINK^® for continuous simulation, and SIMUL8^® for discrete-event simulation) make the mechanics of simulation model building and use relatively straightforward. Stochastic simulation risk models have been developed for business, engineering, biological, social, and economic systems. (Agent-based simulation models have also been developed for complex social and economic systems, but this chapter focuses on continuous simulation.)

Louis Anthony Cox Jr

PART 4 APPLICATIONS AND EXTENSIONS

Chapter 12. Predicting the Effects of Changes: Could Removing Arsenic from Tobacco Smoke Significantly Reduce Smoker Risks of Lung Cancer?

The remainder of the book applies principles from earlier chapters to several challenging quantitative risk assessment (QRA) problems for complex, uncertain, and nonlinear systems. This chapter returns to the problem of predicting how removing a specific constituent (arsenic) from a complex mixture (cigarette smoke) would affect lung cancer risks. This goes beyond the bounding and portfolio QRAs in Chapters 8 and 10 by applying the systems dynamics model in Chapter 11 to obtain explicit quantitative results. Rather than only estimating bounds for the probable changes in consequences, this chapter predicts specific quantitative reductions in risk, contingent on specified assumptions about causal mechanisms. Quantitative sensitivity analysis shows how predicted risk reductions (under stated assumptions) and preventable fractions of risk change as key assumptions are changed.

Louis Anthony Cox Jr

Chapter 13. Simplifying Complex Dynamic Networks: A Model of Protease Imbalance and COPD Dynamic Dose-Response

The risk models in previous chapters have emphasized causal processes described by directed acyclic graphs (Bayesian network models) and by multistage carcinogenic processes in which exposure hastens the net transition rates and increases the net proliferation rates of affected cells. More is required to predict the probable consequences of interventions in feedback-control systems, where changing the levels of controllable inputs can affect not only the specific subprocesses targeted by the interventions, but also the equilibrium levels of other variables throughout the entire interconnected system.

Louis Anthony Cox Jr

Chapter 14. Value of Information (VOI) in Risk Management Policies for Tracking and Testing Imported Cattle for BSE

Previous chapters have discussed and illustrated several methods for building quantitative risk assessment (QRA) models for complex and uncertain systems, including systems with dramatically nonlinear responses (such as the COPD risk model in Chapter 13, where sufficient exposure switches the entire system of interacting feedback-control processes from normal to diseased behavior). Techniques that have proved useful for QRA modeling of complex, uncertain, and potentially nonlinear systems include the following:Information-theory and data-mining algorithms (Chapters 6 and 7) for identifying potential causal relations (including nonlinear and multivariate ones with high-order interactions) in large multivariate data sets.

Louis Anthony Cox Jr

Chapter 15. Improving Antiterrorism Risk Analysis

Several important risk analysis methods now used in setting priorities for protecting U.S. infrastructures against terrorist attacks are based on the formula risk = threat × vulnerability × consequence. This chapter identifies potential limitations in such methods that limit their ability to guide resource allocations to optimize risk reductions. After considering specific examples for the Risk Analysis and Management for Critical Asset Protection (RAMCAP™) framework used by the Department of Homeland Security, we address fundamental limitations of the product formula. These include its failure to adjust for correlations among its components, the nonadditivity of risks estimated using the formula, its inability to use risk-scoring results to allocate defensive resources optimally, and the intrinsic subjectivity and ambiguity of the threat, vulnerability, and consequence numbers.

Louis Anthony Cox Jr

Chapter 16. Designing Resilient Telecommunications Networks

How can telecommunications networks be designed to withstand deliberate attacks by intelligent agents, possibly working in teams? This chapter continues discussing quantitative risk assessment (QRA) for systems with intelligent adversaries by reviewing progress in methods for designing communications networks that are resilient to attacks – that is, that are able to quickly and automatically reroute traffic around affected areas to maintain communications with little or no interruption. Current network architectures, routing and restoration protocols, and design methods already suffice to protect networks against the loss of any single link or node, so the main focus for defending against deliberate attacks is on the design of networks that can reroute traffic even when multiple simultaneous failures occur.

Louis Anthony Cox Jr

Backmatter

Titel: Risk Analysis of Complex and Uncertain Systems
verfasst von: Louis Anthony Cox
Verlag: Springer US
Electronic ISBN: 978-0-387-89014-2
Print ISBN: 978-0-387-89013-5
DOI: https://doi.org/10.1007/978-0-387-89014-2