This paper reviews statistical methods for analyzing output data from computer simulations. First, it focuses on the estimation of steady-state system parameters. The estimation techniques include the replication/deletion approach, the regenerative method, the batch means method, and methods based on standardized time series. Second, it reviews recent statistical procedures to find the best system among a set of competing alternatives.

Proceedings of the 2000 winter simulation conference

A simulation model is composed of inputs and logic; the inputs represent the uncertainty or randomness in the system, while the logic determines how the system reacts to the uncertain elements. Simple input models, consisting of independent and identically distributed sequences of random variates from standard probability distributions, are included in every commercial simulation language. Software to fit these distributions to data is also available. In this tutorial we describe input models that are useful when the input modeling problem is more complex.

/pdf/input-modeling-tools-for-complex-problems-4y3ubcaykh.pdf

Input modeling tools for complex problems

Simulation modellers frequently face a choice between fidelity and variety in their input scenarios. Using an historical trace provides only one realistic scenario. Using the input modelling facilities in commercial simulation software may provide any number of unrealistic scenarios. We ease this dilemma by developing a way to use the moving blocks bootstrap to convert a single trace into an unlimited number of realistic input scenarios. We do this by setting the bootstrap block size to make the bootstrap samples mimic independent realizations in terms of the distribution of distance between pairs of inputs. We measure distance using a new statistic computed from zero crossings. We estimate the best block size by scaling up an estimate computed by analysing subseries of the trace.

https://link.springer.com/content/pdf/10.1057%2Fpalgrave.jors.2601251.pdf

Generation of simulation input scenarios using bootstrap methods

In Gross and Juttijudata (1997) a single node, G/G/1 queue was investigated as to the sensitivity of output performance measures, such as the mean queue wait, to the shape of the interarrival and service distributions selected. Gamma, Weibull, lognormal and Pearson type 5 distributions with identical first and second moments were investigated. Significant differences in output measures were noted for low to moderate traffic intensities (offered load, /spl rho/), in some cases, even as high as 0.8. We continue this type of investigation for two types of queueing networks, namely two versions of a two-node call center, to see if network mixing might reduce the sensitivity effect.

/pdf/sensitivity-of-output-performance-measures-to-input-3pobl15ebi.pdf

Sensitivity of output performance measures to input distributions in queueing simulation modeling

Most discrete-event simulation models have stochastic elements that mimic the probabilistic nature of the system under consideration. A close match between the input model and the true underlying probabilistic mechanism associated with the system is required for successful input modeling. The general question considered here is how to model an element (e.g., arrival process, service times) in a discrete-event simulation given a data set collected on the element of interest. For brevity, it is assumed that data is available on the aspect of the simulation of interest. It is also assumed that raw data is available, as opposed to censored data, grouped data, or summary statistics. This example-driven tutorial examines introductory techniques for input modeling. Most simulation texts (e.g., Law and Kelton 2000) have a broader treatment of input modeling than presented here. Nelson and Yamnitsky (1998) survey advanced techniques.

Input modeling

A simulation model is composed of inputs and logic; the inputs represent the uncertainty or randomness in the system, while the logic determines how the system reacts to the uncertain elements. Simple input models, consisting of independent and identically distributed sequences of random variates from standard probability distributions, are included in every commercial simulation language. Software to fit these distributions to data is also available. In this tutorial we describe input models that are useful when simple models are not.

/pdf/input-modeling-when-simple-models-fail-2abvnicbfk.pdf

James Steinbugl

Papers

Input modeling when simple models fail