A Risk based Approach to Estimate Key Uncertainties

Decisions during the early stages of R&D are often made under substantial uncertainty. Evaluation of R&D alternatives under uncertainty generally does not provide a clear choice that is best under all possible scenarios. For optimal investment of R&D resources, it is important to identify the key uncertainty contributors from a decision maker's perspective. Global sensitivity analysis (GSA) is a tool that can be used to determine key uncertainties that contribute the most to the variance of the bottom-line objective. It is often the case, however, that GSA is not able to distinguish between the uncertainties. Motivated by this, we propose a new tool called conditional – global sensitivity analysis, which further considers the decision-maker's risk preference. The conditional sensitivity measures (cGSAup/cGSAdown) quantify the contributions of different individual uncertainty factors to the upper and lower halves of the distribution function of the objective function. It is argued that the use of cGSAup may appeal to a risk-aversive decision maker as it leads to a lower rate of false acceptance decisions at the expense of a higher rate of false rejection decisions, whereas the use cGSAdown does the opposite.