Statistics and Replications

A single run is one random sample of how your process might unfold. To get a reliable answer, run the model several times as replications and let Output Report summarize the spread of the results.

Showing more than the average

When a model runs with more than one replication, a Statistics control in the report, which reads Statistics (N) with N the number switched on, lets you show more than the average for each figure. Average is always on. You can add:

• Std. Dev., the standard deviation, and Min / Max, to see how much a result varied from run to run.
• Skewness, to see whether the results lean to one side of the average.
• 90% C.I., 95% C.I. and 99% C.I.: a confidence interval, a range you can be that confident the true average falls within.

Each statistic you switch on shows as a sub-row under the figure when you expand it, and you can expand a row to see the value from every replication too. The statistics are not shown on the Bottlenecks, Impact Analysis and Queue Diagnostics tabs. A narrow interval means a precise estimate; if it is too wide to trust, run more replications (see Confidence Interval and the Central Limit Theorem).