Sensitivity analysis allows you to run a model multiple times to see the range of behavior that can be generated under different assumptions. Broadly speaking, there are two kinds of variation that are of interest. The first, often referred to as replication, is used with models containing stochastic elements to see how much variation there might be in the results just because of chance (different drawings from the same random distributions). The second is the variation of behavior that can result from having different initial conditions and constants; this can be done on models with or without any stochastic elements. The term "sensitivity analysis" is often used to mean the second of these, but it really incorporates both.
To set up a sensitivity analysis, you need to specify the number of runs you want and what, if any, changes will be made to constants and initial stock values in the model. The number of runs can be small, in the 5 to 10 range, or quite large. The advantages of a small number of runs are speed and ease of review; they can be completed relatively quickly and you can get a good understanding of the results using Comparative Tables and Graphs. A large number of runs (hundreds, thousands, or more) provides a more complete sampling of input ranges and, most importantly, a more accurate measurement of the range of likely outcomes. If you want to look at outcome distributions, a large number of simulations will be needed.
Continue with Performing Sensitivity Analysis.