AbstractThe uses of simulation are virtually endless. In this article we consider only a few. The first is the use of simulation to take a data set and generate pseudo data points which are essentially stochastic combinations of points in the data set. This SIMDAT application is model free. The second is the use of simulation to obtain data‐based estimates of a stochastic process model. In order to estimate model parameters, we simulate assuming parameter values pseudo data points. Using goodness‐of‐fit measures we can then adjust the parameter values to bring the resulting pseudo data into maximal concordance with the actual data. This SIMEST algorithm is highly a model dependent. Since the time of Poisson, stochatic processes are modeled in the forward (temporal) direction. Fisher's maximum likelihood is a temporally backwards technique of parameter estimation. Thus, for example, if we are looking at a data set of the times of detection of ‘secondary’ tumors, we need to consider all the possible pathways for genesis of these tumors, and there are an infinite number of these. A ‘secondary tumor’ may have been generated from a primary over a continuum of times. It might have been generated from a secondary tumor over a continuum of times. It might actually be a new primary, having been generated from a systemic failure of the immune system. In this paper, we give an example of the forwards SIMEST estimation procedure. SIMEST has been successfully used to generate pseudo data based on an assumption of parameter values and then compared with clinical data via Pearson's goodness of fit. In the area of computational finance, we briefly consider the work which, using simulation‐based data analysis, challenges the efficient market hypothesis and looks at data‐based portfolio alternatives. Copyright © 2010 John Wiley & Sons, Inc.This article is categorized under: Statistical Models > Simulation Models