The output analysis module provides standard statistical analysis tools
for analyzing simulation outputs. Independent, identically-distributed
(IID) samples are analyzed with the
IndependentSample. Time series
(including those with initialization bias and autocorrelation) can be
- The confidence interval provides an upper and lower estimate on a given output, whether that output is an independent, identically-distributed sample or time series data.
- The independent sample is for independent, identically-distributed (IID) samples, or where treating the data as an IID sample is determined to be reasonable. Typically, this will be non-time series data - no autocorrelation. There are no additional requirements on the data beyond being IID. For example, there are no normality assumptions. The
SteadyStateOutputstructs are available for non-IID output analysis.
- Steady-state simulations are useful when the initial conditions and/or final conditions of a simulation are not well-known or not of interest. Steady-state simulation is interested in the long-run behavior of the system, where initial condition effects are negligible. Steady-state simulation analysis is primarily concerned with initialization bias (bias caused by setting initial conditions of the simulation) and auto-correlation (the tendency of a data point in a time series to show correlation with the latest, previous values in that time series). When the interest is a steady-state simulation output, standard simulation design suggests the use of only a single simulation replication.
- Terminating simulations are useful when the initial and final conditions of a simulation are known, and set deliberately to match real world conditions. For example, a simulation spanning a 9:00 to 17:00 work day might use the terminating simulation approach to simulation experiments and analysis. These initial and final conditions are known and of interest.