Techniques for generic probabilistic inversion

Probabilistic inversion of expert assessments to inform projections about Antarctic ice sheet responses

The response of the Antarctic ice sheet (AIS) to changing global temperatures is a key component of sea-level projections. Current projections of the AIS contribution to sea-level changes are deeply uncertain. This deep uncertainty stems, in part, from (i) the inability of current models to fully resolve key processes and scales, (ii) the relatively sparse available data, and (iii) divergent expert assessments. One promising approach to characterizing the deep uncertainty stemming from divergent expert assessments is to combine expert assessments, observations, and simple models by coupling probabilistic inversion and Bayesian inversion. Here, we present a proof-of-concept study that uses probabilistic inversion to fuse a simple AIS model and diverse expert assessments. We demonstrate the ability of probabilistic inversion to infer joint prior probability distributions of model parameters that are consistent with expert assessments. We then confront these inferred expert priors with instrumental and paleoclimatic observational data in a Bayesian inversion. These additional constraints yield tighter hindcasts and projections. We use this approach to quantify how the deep uncertainty surrounding expert assessments affects the joint probability distributions of model parameters and future projections.

Insight into "Calculated Risk": An Application to the Prioritization of Emerging Infectious Diseases for Blood Transfusion Safety

Increasing identification of transmissions of emerging infectious diseases (EIDs) by blood transfusion raised the question which of these EIDs poses the highest risk to blood safety. For a number of the EIDs that are perceived to be a threat to blood safety, evidence on actual disease or transmission characteristics is lacking, which might render measures against such EIDs disputable. On the other hand, the fact that we call them "emerging" implies almost by definition that we are uncertain about at least some of their characteristics. So what is the relative importance of various disease and transmission characteristics, and how are these influenced by the degree of uncertainty associated with their actual values? We identified the likelihood of transmission by blood transfusion, the presence of an asymptomatic phase of infection, prevalence of infection, and the disease impact as the main characteristics of the perceived risk of disease transmission by blood transfusion. A group of experts in the field of infectious diseases and blood transfusion ranked sets of (hypothetical) diseases with varying degrees of uncertainty associated with their disease characteristics, and used probabilistic inversion to obtain probability distributions for the weight of each of these risk characteristics. These distribution weights can be used to rank both existing and newly emerging infectious diseases with (partially) known characteristics. Analyses show that in case there is a lack of data concerning disease characteristics, it is the uncertainty concerning the asymptomatic phase and the disease impact that are the most important drivers of the perceived risk. On the other hand, if disease characteristics are well established, it is the prevalence of infection and the transmissibility of the disease by blood transfusion that will drive the perceived risk. The risk prioritization model derived provides an easy to obtain and rational expert assessment of the relative importance of an (emerging) infectious disease, requiring only a limited amount of information. Such a model might be used to justify a rational and proportional response to an emerging infectious disease, especially in situations where little or no specific information is available.

Identifying and measuring stakeholder preferences for disease prioritisation: A case study of the pig industry in Australia

We describe stakeholder preference modelling using a combination of new and recently developed techniques to elicit criterion weights to incorporate into a multi-criteria decision analysis framework to prioritise exotic diseases for the pig industry in Australia. Australian pig producers were requested to rank disease scenarios comprising nine criteria in an online questionnaire. Parallel coordinate plots were used to visualise stakeholder preferences, which aided identification of two diverse groups of stakeholders - one group prioritised diseases with impacts on livestock, and the other group placed more importance on diseases with zoonotic impacts. Probabilistic inversion was used to derive weights for the criteria to reflect the values of each of these groups, modelling their choice using a weighted sum value function. Validation of weights against stakeholders' rankings for scenarios based on real diseases showed that the elicited criterion weights for the group who prioritised diseases with livestock impacts were a good reflection of their values, indicating that the producers were able to consistently infer impacts from the disease information in the scenarios presented to them. The highest weighted criteria for this group were attack rate and length of clinical disease in pigs, and market loss to the pig industry. The values of the stakeholders who prioritised zoonotic diseases were less well reflected by validation, indicating either that the criteria were inadequate to consistently describe zoonotic impacts, the weighted sum model did not describe stakeholder choice, or that preference modelling for zoonotic diseases should be undertaken separately from livestock diseases. Limitations of this study included sampling bias, as the group participating were not necessarily representative of all pig producers in Australia, and response bias within this group. The method used to elicit criterion weights in this study ensured value trade-offs between a range of potential impacts, and that the weights were implicitly related to the scale of measurement of disease criteria. Validation of the results of the criterion weights against real diseases - a step rarely used in MCDA - added scientific rigour to the process. The study demonstrated that these are useful techniques for elicitation of criterion weights for disease prioritisation by stakeholders who are not disease experts. Preference modelling for zoonotic diseases needs further characterisation in this context.

Global sensitivity analysis of a wave propagation model for arm arteries

Wave propagation models of blood flow and blood pressure in arteries play an important role in cardiovascular research. For application of these models in patient-specific simulations a number of model parameters, that are inherently subject to uncertainties, are required. The goal of this study is to identify with a global sensitivity analysis the model parameters that influence the output the most. The improvement of the measurement accuracy of these parameters has largest consequences for the output statistics. A patient specific model is set up for the major arteries of the arm. In a Monte-Carlo study, 10 model parameters and the input blood volume flow (BVF) waveform are varied randomly within their uncertainty ranges over 3000 runs. The sensitivity in the output for each system parameter was evaluated with the linear Pearson and ranked Spearman correlation coefficients. The results show that model parameter and input BVF uncertainties induce large variations in output variables and that most output variables are significantly influenced by more than one system parameter. Overall, the Young's modulus appears to have the largest influence and arterial length the smallest. Only small differences were obtained between Spearman's and Pearson's tests, suggesting that a high monotonic association given by Spearman's test is associated with a high linear corelation between the inputs and output parameters given by Pearson's test.

Using expert judgment to estimate marine ecosystem vulnerability in the California Current

As resource management and conservation efforts move toward multi-sector, ecosystem-based approaches, we need methods for comparing the varying responses of ecosystems to the impacts of human activities in order to prioritize management efforts, allocate limited resources, and understand cumulative effects. Given the number and variety of human activities affecting ecosystems, relatively few empirical studies are adequately comprehensive to inform these decisions. Consequently, management often turns to expert judgment for information. Drawing on methods from decision science, we offer a method for eliciting expert judgment to (1) quantitatively estimate the relative vulnerability of ecosystems to stressors, (2) help prioritize the management of stressors across multiple ecosystems, (3) evaluate how experts give weight to different criteria to characterize vulnerability of ecosystems to anthropogenic stressors, and (4) identify key knowledge gaps. We applied this method to the California Current region in order to evaluate the relative vulnerability of 19 marine ecosystems to 53 stressors associated with human activities, based on surveys from 107 experts. When judging the relative vulnerability of ecosystems to stressors, we found that experts primarily considered two criteria: the ecosystem's resistance to the stressor and the number of species or trophic levels affected. Four intertidal ecosystems (mudflat, beach, salt marsh, and rocky intertidal) were judged most vulnerable to the suite of human activities evaluated here. The highest vulnerability rankings for coastal ecosystems were invasive species, ocean acidification, sea temperature change, sea level rise, and habitat alteration from coastal engineering, while offshore ecosystems were assessed to be most vulnerable to ocean acidification, demersal destructive fishing, and shipwrecks. These results provide a quantitative, transparent, and repeatable assessment of relative vulnerability across ecosystems to any ongoing or emerging human activity. Combining these results with data on the spatial distribution and intensity of human activities provides a systematic foundation for ecosystem-based management.

Updating parameters of the chicken processing line model

A mathematical model of chicken processing that quantitatively describes the transmission of Campylobacter on chicken carcasses from slaughter to chicken meat product has been developed in Nauta et al. (2005). This model was quantified with expert judgment. Recent availability of data allows updating parameters of the model to better describe processes observed in slaughterhouses. We propose Bayesian updating as a suitable technique to update expert judgment with microbiological data. Berrang and Dickens's data are used to demonstrate performance of this method in updating parameters of the chicken processing line model.

Probabilistic inversion in priority setting of emerging zoonoses

This article presents methodology of applying probabilistic inversion in combination with expert judgment in priority setting problem. Experts rank scenarios according to severity. A linear multi-criteria analysis model underlying the expert preferences is posited. Using probabilistic inversion, a distribution over attribute weights is found that optimally reproduces the expert rankings. This model is validated in three ways. First, consistency of expert rankings is checked, second, a complete model fitted using all expert data is found to adequately reproduce observed expert rankings, and third, the model is fitted to subsets of the expert data and used to predict rankings in out-of-sample expert data.

Regulating under uncertainty: newsboy for exposure limits

Setting action levels or limits for health protection is complicated by uncertainty in the dose-response relation across a range of hazards and exposures. To address this issue, we consider the classic newsboy problem. The principles used to manage uncertainty for that case are applied to two stylized exposure examples, one for high dose and high dose rate radiation and the other for ammonia. Both incorporate expert judgment on uncertainty quantification in the dose-response relationship. The mathematical technique of probabilistic inversion also plays a key role. We propose a coupled approach, whereby scientists quantify the dose-response uncertainty using techniques such as structured expert judgment with performance weights and probabilistic inversion, and stakeholders quantify associated loss rates.