Description

A methodology which investigates a model by generating random numbers and observing the changes in the output. It is usually applied when a problem cannot be solved analytically.

Software

There is a large number of packages available to perform such an analysis, including stand-alone packages and Excel add-ins, such as @RISK, Crystal Ball, SimLab? and many others. It can be easily performed from any statistical software package such as Matlab or R. Many of these packages also allow for a more effective sampling design to for example distribute the random numbers evenly and avoid clusters.

Advantages

• Forces explicit acknowledgement of all sources of uncertainty All sources of uncertainty have to be stated.

• Can take account of any distribution and correlation The method is not restricted to a specific set of distributions and can deal with all forms of correlation

• Can include other sources of error The methodology can virtually include all sources of uncertainty

• Can be applied to complex models model complexity is not a limiting factor as in so many other methodologies.

Disadvantagess

• Computationally intensive The methodology is computational intensive and become infeasible with models which have a long run time or too many sources of uncertainty.

References and Further reading

EPA, Risk assessment Forum, Guiding Principles for Monte Carlo Analysis, EPA/630/97/001, 1997. 1997. 158

Morgan, B.J., Elements of simulation. 1984: Chapman & Hall. 159

Ripley, B.D., Stochastic simulation. 1978: Wiley160