A unifying approach to risk and sensitivity analyses

There is a decision paradox that can prevent strategic decisions from being made both quickly and on a sound basis. A method which combines the advantageous elements of probabilistic risk and sensitivity analyses is proposed to overcome this paradox. This unified approach extends a previously developed method of stochastic sensitivity analysis to generate approximations of the outcome distribution parameters. The results are validated by comparison against analytical solutions in the case of simple models and against simulation results for more complex models. The research findings show that the new method gives standard deviation estimates which are comparable to Monte Carlo simulation estimates, even though the new estimates may be biased. The cumulative distributions estimated by the approach are generally similar to the simulated distributions so that the unified approach and simulation lead to consistent decisions being made. The approach is robust and produces good results when the input distributions are badly skewed and when there is correlation among the input variables. The new method has a potentially large time advantage over simulation when changes in the input distributions or model must be investigated.