Usage

Most of the time and for a given calculation budget, a compromise needs to be found between the number of runs to execute (that depend on the CPU time required for each run) and the statistical quality of the Monte-Carlo sampling method. This is illustrated in the following figure which compares Random, LH and OLH techniques:

Figure 91: Sampling method selection

The important points to remember are:

• For a given number of runs, the random sampling, sometimes called "brute force", is a computationally-cheap strategy that provides no specific guarantees in terms of statistical quality. It is useful for quick and simple robustness or variability assessment.

• LH provides a better representation of real-world variability and can be seen as a good compromise in most cases since the sampling generation, even with better space-filling, is not very costly.

• OLH provides a very accurate representation of the real-world variability but the counterpart is that it can be very costly to generate (an optimization procedure is required).

See below for more details on statistical quality and LH / OLH sampling methods.