SIMTBED a graphical test bed for analyzing and reporting the results of a statistical simulation experiment

Abstract

A graphical test bed in which the results of simulation experiment can be reported and analyzed is described. The test bed is based o the regression adjusted graphics and estimation methodology developed by Heidelberger and Lewis (Ref. 1) for regenerative simulation. From the graphics and associated numeric, the experimenter can summarize and see simultaneously relative properties, such as bias, normality and standard deviation, of several estimators of a characteristic of a population for up to 8 sample sizes. The evolution of these properties with sample size is also displayed. The graphics is supported on a lien printer to make it and the program portable. The technique is illustrated by examples concerning the effects of changes in data distribution on the behavior of the lag one serial correlation coefficient, the estimation of the shape parameter of Gamma random variables and a comparison of different methods (jackknife, bootstrap) for estimating the standard error of an estimator.