Caroline:
After your presentation in Glasgow, I realised that I should ask you to collaborate on a small paper that illustrates the use of c-boxes to express the uncertainty of fault tree probabilities obtained by expert elicitation.
We discussed this casually on the train, and you suggested you could, but I know you are very busy. So let us talk about what it might take and whether you can participate. Perhaps we could steal some time during the holiday to talk or even actually do something.
This evening, by sheer chance, I stumbled on to some long-lost files that we used to undertake expert elicitations of some geneticists and ecologists for a fault tree analysis of the risks of introducing mosquitoes in Africa that are modified with the homing endonuclease gene (HEG). This was the example I was hoping to use for the paper.
The paper we should write is very simple. It argues that a new approach is needed for characterising the leaves of fault trees, especially when those leaves represent rare events whose probabilities are estimated by fallible expert elicitations. I think it should use the HEG risk assessment as a numerical example, but we could include a second example more relevant for, say, offshore facilities you worry about.
I think the following are the links to variously relevant resources:
the files (attached) from the HEG risk assessment elicitation exercise including calculation of one of the HEG risk assessment's subproblems,
a presentation (attached) I gave at the 2018 EPSRC HubNet Risk Day in Manchester that contextualises the engineering problem,
the c-box website https://sites.google.com/site/confidenceboxes/ and, in particular,
the simulations showing their statistical 'performance' https://sites.google.com/site/confidenceboxes/simulations/binomial-distribution,
three websites (not yet integrated) on estimating probability of rare events
https://sites.google.com/site/beyondthebagofmarbles/rare-and-unseen-events
https://sites.google.com/site/computingwithconfidence/say-1/zeronumeratorproblem
the website on c-boxes in reliability theory (including Jackson's incoherent tree that you hate) https://sites.google.com/site/reliabilityuncertainty/,
the HEG risk assessment site https://sites.google.com/site/mosquitorisk/,
Keith's elicited probabilities on other subproblems which are among my emails,
an old ICOSSAR proceedings paper on an uncertainty penalty for EE-ed probabilities,
the website on EBC for translating the results back into an expression of the form "k out of n" https://sites.google.com/site/equivalentbinomialcount/, and
our EE website https://sites.google.com/site/liverpoolexpertelicitation/.
We might want to ask Keith Hayes to be a coauthor, in which case authorship should be Morais, Hayes and Ferson. Maybe we need others too. I think that Alex Wimbush would be interested.
Let me know if you have time in the next couple of weeks to discuss this briefly.
Cheers,
Scott