Changing the Random Seed

Each Probability Distribution has the RandomSeed keyword, which generates a different pseudo-random sequence of values for each Probability Distribution.  Often, a simulation model is executed multiple times, called ‘replications’, using different random seeds to determine the statistical accuracy of the output parameters.

One way to achieve multiple replications is to change the RandomSeed for each Probability Distribution.  However, this can be impractical when large numbers of distributions are used in a model.  The GlobalSubstreamSeed keyword for the Simulation object simplifies this process.  Changing the input for this keyword causes all Probability Distributions in the model to generate a different random sequence.  In effect, the seed for each Probability Distribution is the combination of the inputs to the RandomSeed and GlobalSubstreamSeed keywords.

When multiple objects behave according to a Probability Distribution with the same characteristics, it is advisable to have a unique instance of the Probability Distribution for each object.  This avoids any cross-correlation effects arising from the order in which a single distribution may be sampled.