Configuring the Sampling Technique and Execution Options

You use the General tab to select the sampling technique and to define the tuning parameters associated with the technique. In addition, you can define options that Isight uses when the component executes.

  1. From the Monte Carlo Component WebTop Editor, click the General tab.

  2. Choose one of the options from the Sampling Technique list.

    Option Description
    Descriptive Sampling This method is the recommended method. Fewer simulations are required to estimate the statistical characteristics of the system behavior.
    Simple Random Sampling This method is the default selection and is the most commonly used sampling method. Sample points for simulation are generated randomly and independently.
    Sobol Sampling This method generates numbers as binary fractions of appropriate length from a set of special binary fractions.

    The sampling technique’s options appear directly below the Sampling Technique list. Information about the technique appears in the Sampling Technique Description area on the right side of the editor.

    For more detailed information on these sampling techniques, see About the Sampling Techniques in the Isight Component Guide.

  3. Set the following options in the Sampling Technique Options area:

    Option Description
    Number of Simulations Determine the number of simulations to be executed. The default number is 1000 for simple random sampling and 100 for descriptive sampling. A higher number of simulations specified generally results in an increase in the accuracy level of the statistical predictions.
    Use a fixed seed Type the value of the random seed in the corresponding text box. If this option is not selected, Isight determines the seed randomly.

    Note: A fixed seed can also be set and used for the entire model. If a fixed seed is set for the model and also defined separately in the Monte Carlo component, Isight uses the one defined in the Monte Carlo component. If a fixed seed is set for the model and this option is not selected, the sequence of random numbers is based on the model’s random seed (and, thus, is still reproducible). For information about setting a fixed seed, see Setting the Execution Options in the Isight User’s Guide.

  4. Set the following options in the Execution Options area:

    Option Description
    Execute Monte Carlo sample points in parallel Select this option if you want the sample points defined by the Monte Carlo simulation to execute in parallel. The convergence check interval may limit the number of sample points that are executed in parallel to batches that are the size specified for this convergence check interval.
    After execution, reset to mean value point and run With this option selected (default), after execution of all Monte Carlo simulation points, the random variables are set to their mean values and the Monte Carlo subflow executes one additional time. In this case, the Monte Carlo parameters (inputs and outputs) are left at the values calculated for this “mean value point.” If this option is not selected, the Monte Carlo parameters are left at the values associated with the last (random) Monte Carlo simulation point after execution.
    Convergence Check Interval This integer value specifies the frequency at which the convergence of the Monte Carlo response statistics (low order—mean and standard deviation) are checked. The default setting requires the convergence of response statistics to be checked after every 25 sample points. Decreasing this value may result in premature convergence; because the sample points are random, or in random order/combinations, sufficient sample points are needed to determine if the response statistics are estimated accurately. If this value is set to zero, then the convergence check is disabled.
    Convergence Tolerance This value is the Monte Carlo termination criterion, which specifies the fractional change in response mean and standard deviation that will allow the simulation to be terminated before reaching the specified number of simulations. For example, if you want the simulation process to stop if the change in the response statistics goes below 5%, enter a value of 0.05 in the corresponding text box. Convergence is checked after each set of simulations defined by the Convergence Check Interval (by default, after each set of 25 simulations). The default Convergence Tolerance is 0.001. If this value is set to zero, then the convergence check is disabled.

  5. Click Import to import all or part of the Monte Carlo component configuration from a file. Click Export to export the entire configuration to a file.

    This option provides you with an easy mechanism for quickly configuring the component using information from an external source without manually entering all the information in the component editor. For more information, see About Importing and Exporting Configuration Files in the Isight Component Guide.

  6. Click Update Component to save your changes.