# Simulation as a data-analytic tool for complex systems

#### by Dr Larry Weldon

Institution: Dept of Statistics and Actuarial Science, Simon Fraser University
Date: Thu 2nd June 2005
Time: 12:30 PM
Location: Theatre 1, Old Geology

Abstract: Simulation has been used extensively in teaching and research in statistics. (Mills (2002)). Its principal use for statistics has been either to demonstrate statistical phenomena or to calculate some property of a model that is not available through direct analysis. More recently, The role of simulation for data analysis itself has been promoted recently by advocates of resampling as a general-purpose tool for statistical analysis. (Simon(1993), Good(2001)). This use of simulation as a tool of inference rather than a tool of statistical theory is fairly new. In this paper I wish to draw attention to another use of simulation for inference. Data in this setting is a complex phenomenon, like a highway traffic accordion, sports league ranking phenomena, retail sales under stock constraints, or the relationship between lifelong health status of individuals and instantaneous hospital stay status. In models like these, some of the connections among variables can be modeled, although analysis of the model by mathematics may not be feasible. In situations like these, simulation models can be calibrated to match the observed phenomenon. Then the impact of changing parameter values can be assessed without changing the real-world process.

The talk will outline four examples of this kind of analysis: calibrate a model to match some observed outcome features, and then use the model to help extract more information from the data. R-programs are used to demonstrate the simulation results, although the models are simple enough that any statistical package could do the job. The important role of graphics in analysis involving simulation is emphasized, with implications for course content.

Mills, Jamie D. (2002) Using Computer Simulation Methods to Teach Statistics: A Review of the Literature. Journal of Statistics Education 10(1). (www.amstat.org/publications/jse/v10n1/mills.html)

Simon, J. (1993) Resampling: The "New Statistics". Wadsworth.

Good, P. I. (2001) Resampling Methods : a practical guide to data analysis. Second Edition. Birkhauser, Boston.