Monte Carlo Simulation Checklist

This is a Monte Carlo simulation checklist to ensure proper model set-up when working with Simulation Master.  Unless stated otherwise, each item refers to both the standard and premium editions.

Monte Carlo Simulation Checklist

  • If you want to record random variable values in the simulation data sheet, the random variables must be in the same worksheet as the model output cell.
  • To be recorded in the simulation data sheet, a cell's formula must begin with "=RV".  For random variables with additional math operations, such as =2*RVNORMAL(10,2), enclose the formula in the RVUSERDIST function.  Example: =RVUSERDIST(2*RVNORMAL(10,2)).
  • User defined random variables created using the RVUSERDIST function must evaluate to a numeric value.  If a Boolean result is desired, use a numeric coding scheme such as 0 = FALSE and 1 = TRUE.
  • Spreadsheet calculation: If random variable cells are in a different worksheet or workbook than the model output cell, select calculate all workbooks prior to running the simulation.
  • If using rank correlation sampling for input random variables, the correlation matrix and all correlated random variables must be on the same worksheet as the model output cell.
  • (Premium Edition) All discrete or optimization decision variables must be in the same worksheet as the model output cell.
  • (Premium Edition)  When using optimization, all cell constraints and their referenced values must be in the same worksheet as the model output cell.
  • The model output cell must evaluate to a numeric value.  If a Boolean result is desired, use a numeric coding scheme such as 0 = FALSE and 1 = TRUE.

Optimization Decision Variable Bounds (Premium Edition)

To improve optimization efficiency, keep minimum and maximum bounds as close to realistic values as possible.  Keeping the decision variables within reasonable bounds limits the search space which allow the software a better chance to find an optimal solution or takes fewer optimization iterations to find an optimal solution.

For example, if a decision variable corresponds to length, the minimum should be at least zero since we cannot have a negative length.  The maximum should be set at a value the would realistically be the maximum possible value.

Equality vs. Inequality Constraints (Premium Edition)

When using constraints during optimization, using inequality constraints are more efficient than equality constraints.  If possible, it is preferable to use an inequality constraint than an equality constraint.

For example, if there is a constraint that the 5th percentile of the simulation must be 0 we could use an equality constraint with the 5th percentile equal to 0.  If we can live with the 5th percentile greater than 0, a better approach would be to use an inequality constraint with the 5th percentile greater than or equal to 0.