Score averaging for alien species risk assessment: a probabilistic alternative

In many biological risk assessment problems and specifically those involving invasive alien species and quarantine pests, ordinal scoring systems are used as a basis for risk quantification. Frequently, the problem is disaggregated into a series of components to allow a more structured and systematic evaluation. Each component is then scored and an overall measure of risk calculated from the set of scores. No satisfactory method exists to compute an overall measure of risk and most practitioners have resorted to score-averaging. Here we describe an approach involving the linear mapping of scores to notional probabilities to allow the use of conditional probability to derive an overall measure. Initial results give better discrimination between high and low risk cases and closer correspondence with independent expert judgement than does score-averaging.