Sampling Settings

Figure 1 - Simulation Settings Window - Sampling Tab

The Sampling tab of the Simulation Settings window (Figure 1, right) includes two options groups:

  • Random Numbers - Configurations for the methods for Sampling and Random Number Generation @RISK will utilize during simulation.
  • Other Options - Settings for how @RISK will collect distribution sample values, enabling or disabling Smart Sensitivity Analysis, and when to update any @RISK Statistic functions that exist in the model.

See Iterations and Simulations for more information on sampling

Random Number Options

These settings control the methods by which @RISK will generate random values for distribution samples for all @RISK input functions.

Please note: these options can have a significant impact on the results of a simulation! They should only be modified if the effects of any changes are well understood.

  • Sampling Type - Set the type of sampling used during a simulation run. See Sampling Methods for more information on how these two methods function and differ.
    • Latin Hypercube - This is the recommended setting. Selects a stratified sampling; this method more accurately recreates the sample values for a probability distribution (or takes fewer iterations to achieve similar accurately).
    • Monte Carlo - Selects standard Monte Carlo sampling.
  • Generator - Set the random number generator utilized during a simulation run. For details on the Generator types, see Random Number Generation.
  • Initial Seed - The initial seed value used by the Generator; a seed is a number that initializes the selection of numbers by a random number generator; given the same seed number, a random number generator will generate the same series of random numbers each time a simulation is run.
    • Choose Randomly - Default Setting. @RISK will use a randomly selected number as the Seed value.
    • Fixed - When this option is selected, a field will be activated. Enter an integer as the Seed value
    • Using a Fixed seed is a useful tool for collaboration! Fixed seeds not only control variability, they also guarantee reproducibility of simulation results; given the same seed, a random number generator (and therefore, @RISK distribution functions) will generate the same series of random numbers each time a simulation is run.

  • Multiple Simulations - Select how multiple simulations will generate Seed values.
    • All Use Same Seed - Every simulation will utilize the same Seed value as configured in the 'Initial Seed' setting.
    • Use Different Seeds - Each simulation will generate its own Seed value.
    • If a Fixed Seed is used for the Initial Seed and Use Different Seeds option is selected, each simulation will use a different seed, but the same sequence of seed values will be used each time the simulation run is executed. Similar to using a Fixed Seed, this guarantees reproducibility between simulation runs!

Other Options

The 'Other Options' settings determine how (or if) @RISK collects and stores distribution sample values, among other settings related to distribution samples. Collection of distribution sample values and their statistics aids in analysis through Simulation Sensitivities and Simulation Scenarios by preserving the values upon which various other statistics can be generated. Additionally, saved sample values can be viewed in the Simulation Data window (e.g. for debugging purposes).

  • Collect Distribution Samples - Set which distributions should have their sample values collected.
    • All - Collect samples on all distribution functions in the model.
    • Inputs Marked With Collect - Only collect samples from those distribution functions that have been specifically configured to be included in collection. See Configure Distribution for more information on configuring individual inputs for collection.
    • None - No collection of distribution samples.
  • Smart Sensitivity Analysis - Enable or disable Smart Sensitivity Analysis; see Smart Sensitivity Analysis for more information.
  • Update Statistic Functions - Configure when @RISK will update any Statistic functions that have been inserted into the model.
    • Each Iteration - Any statistic functions (RiskMean, RiskStdDev, etc.) in the model will be recalculated after every iteration.
    • In most cases, statistics do not need to be updated until the end of a simulation; however, if a model requires regular updates to these statistics, such as when a custom convergence calculation is included, this can be enabled.

    • Please note: updating Statistic functions after each iteration can add significant runtime to large or complicated models.

    • At the End of Each Simulation - Statistic functions will be updated at the end of a simulation. If multiple simulations are executed during a run, statistic functions will be updated each time a simulation completes.