Modelling Under Risk and Uncertainty: An Introduction to Statistical, Phenomenological and Computational Methods

2.6 Computing an Appropriate Risk Measure or Quantity of Interest and Associated Sensitivity Indices

A large variety of risk measures may therefore be considered: formulating a different probabilistic property such as exceedance probability or variance; distinguishing or not the aleatory and epistemic components of uncertainty in several possible ways; and incorporating more or less mixed deterministic-probabilistic settings. Leaving to Chapter 4 the discussion of their implications regarding the nature of the underlying systems and the purpose of decision-making, let it be noted that the regulatory or practical formulation of the risk or uncertainty assessment may sometimes specify the appropriate risk measure clearly, as illustrated in Table 2.3.

Table 2.3 Formulation of decision criteria and corresponding paradigms.

Formulation of the risk measure or decision criterion	Domain of typical application	Type of risk measure
‘there should be less than 3% uncertainty on the declared value for the output of interest’	Uncertainty analysis or calibration	Variance
‘the physical margin should remain positive in spite of uncertainty, with a probability less than 10−b to be negative’	Structural safety	Exceedance probability
Large Early Release Fraction should keep below a given dose for an expected frequency of 1/10 000 (i.e. averaged over the epistemic uncertainty affecting the frequency value)	Nuclear IPRA	Quantile
‘frequency of failure should be less than 10−b per year, at ...