Applies to: @RISK 5.x and Up

Sensitivity Simulation with RiskSimtable()

Background:

Standard sensitivity analysis shows the effect of random uncertainty, but some uncertain inputs are under your control. For these controllable variables, you want to test fixed alternative values (prices, materials, bids, strategies) and compare results. @RISK’s Sensitivity Simulation is designed for this purpose.

A Sensitivity Simulation runs multiple simulations in one run, each using a different set of user-defined values. This is done with the RiskSimtable() function, which supplies a list of values that are used one per simulation.

Example:
=RiskSimtable({100,200,300,400})
Simulation 1 uses 100, simulation 2 uses 200, simulation 3 uses 300, and simulation 4 uses 400.

How to run a Sensitivity Simulation:

1. Enter RiskSimtable() in the cells or formulas for user-controlled variables.

2. Open Simulation Settings and set the Number of Simulations to match the number of values (4 in the example).

3. Run the simulation.

Each simulation runs the same number of iterations, collects the same outputs, and differs only in the RiskSimtable() values. @RISK then processes and displays results for each simulation side by side.

Analyzing results:
Each output produces a separate distribution per simulation. You can compare scenarios using Distribution Summary graphs and the Simulation Summary report to see how results change across alternatives.

Advanced use:
RiskSimtable() arguments can themselves be distributions. This lets you compare different distribution choices (e.g., RiskTriang vs RiskPert vs RiskNormal) across simulations.

Important cautions:

• Use RiskSimtable() for controlled scenario changes, not for random discrete events. For random outcomes, use distribution functions such as RiskDiscrete().

• By default, all simulations use the same random seed so differences are due only to RiskSimtable() values. To change this, enable Multiple Simulations Use Different Seed Values on the Sampling tab.

Sensitivity Simulation is ideal when your model includes both random uncertainty and decisions that you will ultimately choose based on comparative results.

 

Attachments:
SenSim.xls